103. Skull_segmentation_with_Unet_and_autoencoder.ipynb

{"cells":[{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":4663,"status":"ok","timestamp":1665128069251,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"},"user_tz":-360},"id":"I2lGtVUzadL8","outputId":"e084aded-8369-44ac-cd7c-0bcf8f7bd6d0"},"outputs":[{"output_type":"stream","name":"stdout","text":["Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n","Requirement already satisfied: tensorflow in /usr/local/lib/python3.7/dist-packages (2.8.2+zzzcolab20220929150707)\n","Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (3.3.0)\n","Requirement already satisfied: gast>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (0.5.3)\n","Requirement already satisfied: keras-preprocessing>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (1.1.2)\n","Requirement already satisfied: flatbuffers>=1.12 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (22.9.24)\n","Requirement already satisfied: astunparse>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (1.6.3)\n","Requirement already satisfied: tensorflow-io-gcs-filesystem>=0.23.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (0.27.0)\n","Requirement already satisfied: tensorflow-estimator<2.9,>=2.8 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (2.8.0)\n","Requirement already satisfied: protobuf<3.20,>=3.9.2 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (3.17.3)\n","Requirement already satisfied: numpy>=1.20 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (1.21.6)\n","Requirement already satisfied: libclang>=9.0.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (14.0.6)\n","Requirement already satisfied: grpcio<2.0,>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (1.49.1)\n","Requirement already satisfied: typing-extensions>=3.6.6 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (4.1.1)\n","Requirement already satisfied: h5py>=2.9.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (3.1.0)\n","Requirement already satisfied: termcolor>=1.1.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (2.0.1)\n","Requirement already satisfied: wrapt>=1.11.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (1.14.1)\n","Requirement already satisfied: tensorboard<2.9,>=2.8 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (2.8.0)\n","Requirement already satisfied: six>=1.12.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (1.15.0)\n","Requirement already satisfied: absl-py>=0.4.0 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (1.2.0)\n","Requirement already satisfied: setuptools in /usr/local/lib/python3.7/dist-packages (from tensorflow) (57.4.0)\n","Requirement already satisfied: google-pasta>=0.1.1 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (0.2.0)\n","Requirement already satisfied: keras<2.9,>=2.8.0rc0 in /usr/local/lib/python3.7/dist-packages (from tensorflow) (2.8.0)\n","Requirement already satisfied: wheel<1.0,>=0.23.0 in /usr/local/lib/python3.7/dist-packages (from astunparse>=1.6.0->tensorflow) (0.37.1)\n","Requirement already satisfied: cached-property in /usr/local/lib/python3.7/dist-packages (from h5py>=2.9.0->tensorflow) (1.5.2)\n","Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow) (0.4.6)\n","Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow) (3.4.1)\n","Requirement already satisfied: google-auth<3,>=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow) (1.35.0)\n","Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow) (0.6.1)\n","Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow) (1.8.1)\n","Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow) (1.0.1)\n","Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard<2.9,>=2.8->tensorflow) (2.23.0)\n","Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (4.9)\n","Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (0.2.8)\n","Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (4.2.4)\n","Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.9,>=2.8->tensorflow) (1.3.1)\n","Requirement already satisfied: importlib-metadata>=4.4 in /usr/local/lib/python3.7/dist-packages (from markdown>=2.6.8->tensorboard<2.9,>=2.8->tensorflow) (5.0.0)\n","Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard<2.9,>=2.8->tensorflow) (3.8.1)\n","Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard<2.9,>=2.8->tensorflow) (0.4.8)\n","Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (2022.9.24)\n","Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (1.24.3)\n","Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (3.0.4)\n","Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard<2.9,>=2.8->tensorflow) (2.10)\n","Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard<2.9,>=2.8->tensorflow) (3.2.1)\n"]}],"source":["!pip install tensorflow"]},{"cell_type":"markdown","source":["**Acknowledgements:**\n","\n","The codes are taken from\n","[Digital sreeni ](https://www.youtube.com/watch?v=hTpq9lzAb8M&t=1365s) and [Idiot developer ](https://www.youtube.com/watch?v=LkmgdNpcgTM&t=841s) then few changes are made for this project.\n"],"metadata":{"id":"JSkDW75veT94"}},{"cell_type":"markdown","source":["**Building The Model**"],"metadata":{"id":"f1vhFujERYyc"}},{"cell_type":"code","execution_count":5,"metadata":{"executionInfo":{"elapsed":433,"status":"ok","timestamp":1665146620555,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"},"user_tz":-360},"id":"AR8keUmgaoei"},"outputs":[],"source":["from tensorflow.keras.models import Model\n","from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, UpSampling2D, concatenate, Conv2DTranspose, BatchNormalization, Dropout, Lambda\n","from tensorflow.keras.optimizers import Adam\n","from tensorflow.keras.layers import Activation, MaxPool2D, Concatenate\n","\n","#Convolutional block to be used in autoencoder and U-Net\n","def convolution_block(input, num_filters):\n","    x = Conv2D(num_filters, 3, padding=\"same\")(input)\n","    x = BatchNormalization()(x)   #Not in the original network. \n","    x = Activation(\"relu\")(x)\n","\n","    x = Conv2D(num_filters, 3, padding=\"same\")(x)\n","    x = BatchNormalization()(x)  #Not in the original network\n","    x = Activation(\"relu\")(x)\n","\n","    return x\n","\n","#Encoder block: Conv block followed by maxpooling\n","def encoder_block(input, num_filters):\n","    x = convolution_block(input, num_filters)\n","    p = MaxPool2D((2, 2))(x)\n","    return x, p   \n","\n","#Decoder block for autoencoder (no skip connections)\n","def decoder_block(input, num_filters):\n","    x = Conv2DTranspose(num_filters, (2, 2), strides=2, padding=\"same\")(input)\n","    x = convolution_block(x, num_filters)\n","    return x\n","\n","#Encoder will be the same for Autoencoder and U-net\n","#We are getting both conv output and maxpool output for convenience.\n","#we will ignore conv output for Autoencoder. It acts as skip connections for U-Net\n","def build_encoder(input_image):\n","    #inputs = Input(input_shape)\n","\n","    s1, p1 = encoder_block(input_image, 64)\n","    s2, p2 = encoder_block(p1, 128)\n","    s3, p3 = encoder_block(p2, 256)\n","    s4, p4 = encoder_block(p3, 512)\n","    \n","    encoded = convolution_block(p4, 1024) #Bridge\n","    \n","    return encoded\n","\n","#Decoder for Autoencoder ONLY. \n","def build_decoder(encoded):\n","    d1 = decoder_block(encoded, 512)\n","    d2 = decoder_block(d1, 256)\n","    d3 = decoder_block(d2, 128)\n","    d4 = decoder_block(d3, 64)\n","    \n","    decoded = Conv2D(3, 3, padding=\"same\", activation=\"sigmoid\")(d4)\n","    return decoded\n","\n","#Use encoder and decoder blocks to build the autoencoder. \n","def build_autoencoder(input_shape):\n","    input_img = Input(shape=input_shape)\n","    autoencoder = Model(input_img, build_decoder(build_encoder(input_img)))\n","    return(autoencoder)\n","\n","\n","#Decoder block for unet\n","#skip features gets input from encoder for concatenation\n","def decoder_block_for_unet(input, skip_features, num_filters):\n","    x = Conv2DTranspose(num_filters, (2, 2), strides=2, padding=\"same\")(input)\n","    x = Concatenate()([x, skip_features])\n","    x = convolution_block(x, num_filters)\n","    return x\n","\n","#Build Unet using the blocks\n","def build_unet(input_shape):\n","    inputs = Input(input_shape)\n","\n","    s1, p1 = encoder_block(inputs, 64)\n","    s2, p2 = encoder_block(p1, 128)\n","    s3, p3 = encoder_block(p2, 256)\n","    s4, p4 = encoder_block(p3, 512)\n","\n","    b1 = convolution_block(p4, 1024) #Bridge\n","\n","    d1 = decoder_block_for_unet(b1, s4, 512)\n","    d2 = decoder_block_for_unet(d1, s3, 256)\n","    d3 = decoder_block_for_unet(d2, s2, 128)\n","    d4 = decoder_block_for_unet(d3, s1, 64)\n","\n","    outputs = Conv2D(1, 1, padding=\"same\", activation=\"sigmoid\")(d4)  #Binary (can be multiclass)\n","\n","    model = Model(inputs, outputs, name=\"U-Net\")\n","    #print(model.summary())\n","    return model\n","\n","# model1=build_unet((512, 512, 3))\n","# model2=build_autoencoder((512, 512, 3))\n","# print(model1.summary())\n","\n","\n"]},{"cell_type":"markdown","source":["Loading the data"],"metadata":{"id":"XV3O3HBcRe2W"}},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":1096,"status":"ok","timestamp":1665075683702,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"},"user_tz":-360},"id":"w0O8B49_asWs","outputId":"cc31282f-dd82-45e1-eb9d-fb7effcba3e7"},"outputs":[{"output_type":"stream","name":"stderr","text":["100%|██████████| 157/157 [00:00<00:00, 201.24it/s]\n"]}],"source":["import numpy as np\n","import cv2\n","from tensorflow.keras.preprocessing.image import img_to_array\n","from tensorflow.keras.layers import Conv2D, MaxPooling2D, UpSampling2D, Dropout, Input, Conv2DTranspose\n","from tensorflow.keras.models import Sequential\n","from tensorflow.keras.callbacks import ModelCheckpoint, CSVLogger, ReduceLROnPlateau\n","import os\n","from keras.models import Model\n","from matplotlib import pyplot as plt\n","SIZE=256\n","\n","\n","from tqdm import tqdm\n","img_data=[]\n","path1 = '/content/drive/MyDrive/Colab Notebooks/Skull images'\n","files=os.listdir(path1)\n","for i in tqdm(files):\n","    img=cv2.imread(path1+'/'+i,1)   #Change 0 to 1 for color images\n","    img=cv2.resize(img,(SIZE, SIZE))\n","    img_data.append(img_to_array(img))"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":13,"status":"ok","timestamp":1665075683704,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"},"user_tz":-360},"id":"b9ZvNWDqa0u1","outputId":"a3c18be6-b5e3-4db7-a4f0-6a2c647442cd"},"outputs":[{"output_type":"stream","name":"stdout","text":["157\n"]}],"source":["print(len(img_data))"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"UGO-gWF6bCv8"},"outputs":[],"source":["img_array = np.reshape(img_data, (len(img_data), SIZE, SIZE, 3))\n","img_array = img_array.astype('float32') / 255."]},{"cell_type":"markdown","source":["Calling autoencoder model"],"metadata":{"id":"RYKHnh-bRjqa"}},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":1749,"status":"ok","timestamp":1665076645263,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"},"user_tz":-360},"id":"rUVOibs0a3-X","outputId":"e4f304dc-d5fe-43ad-8cbc-6c3a8eaa5fcc"},"outputs":[{"output_type":"stream","name":"stdout","text":["Model: \"model_2\"\n","_________________________________________________________________\n"," Layer (type)                Output Shape              Param #   \n","=================================================================\n"," input_3 (InputLayer)        [(None, 256, 256, 3)]     0         \n","                                                                 \n"," conv2d_38 (Conv2D)          (None, 256, 256, 64)      1792      \n","                                                                 \n"," batch_normalization_36 (Bat  (None, 256, 256, 64)     256       \n"," chNormalization)                                                \n","                                                                 \n"," activation_36 (Activation)  (None, 256, 256, 64)      0         \n","                                                                 \n"," conv2d_39 (Conv2D)          (None, 256, 256, 64)      36928     \n","                                                                 \n"," batch_normalization_37 (Bat  (None, 256, 256, 64)     256       \n"," chNormalization)                                                \n","                                                                 \n"," activation_37 (Activation)  (None, 256, 256, 64)      0         \n","                                                                 \n"," max_pooling2d_8 (MaxPooling  (None, 128, 128, 64)     0         \n"," 2D)                                                             \n","                                                                 \n"," conv2d_40 (Conv2D)          (None, 128, 128, 128)     73856     \n","                                                                 \n"," batch_normalization_38 (Bat  (None, 128, 128, 128)    512       \n"," chNormalization)                                                \n","                                                                 \n"," activation_38 (Activation)  (None, 128, 128, 128)     0         \n","                                                                 \n"," conv2d_41 (Conv2D)          (None, 128, 128, 128)     147584    \n","                                                                 \n"," batch_normalization_39 (Bat  (None, 128, 128, 128)    512       \n"," chNormalization)                                                \n","                                                                 \n"," activation_39 (Activation)  (None, 128, 128, 128)     0         \n","                                                                 \n"," max_pooling2d_9 (MaxPooling  (None, 64, 64, 128)      0         \n"," 2D)                                                             \n","                                                                 \n"," conv2d_42 (Conv2D)          (None, 64, 64, 256)       295168    \n","                                                                 \n"," batch_normalization_40 (Bat  (None, 64, 64, 256)      1024      \n"," chNormalization)                                                \n","                                                                 \n"," activation_40 (Activation)  (None, 64, 64, 256)       0         \n","                                                                 \n"," conv2d_43 (Conv2D)          (None, 64, 64, 256)       590080    \n","                                                                 \n"," batch_normalization_41 (Bat  (None, 64, 64, 256)      1024      \n"," chNormalization)                                                \n","                                                                 \n"," activation_41 (Activation)  (None, 64, 64, 256)       0         \n","                                                                 \n"," max_pooling2d_10 (MaxPoolin  (None, 32, 32, 256)      0         \n"," g2D)                                                            \n","                                                                 \n"," conv2d_44 (Conv2D)          (None, 32, 32, 512)       1180160   \n","                                                                 \n"," batch_normalization_42 (Bat  (None, 32, 32, 512)      2048      \n"," chNormalization)                                                \n","                                                                 \n"," activation_42 (Activation)  (None, 32, 32, 512)       0         \n","                                                                 \n"," conv2d_45 (Conv2D)          (None, 32, 32, 512)       2359808   \n","                                                                 \n"," batch_normalization_43 (Bat  (None, 32, 32, 512)      2048      \n"," chNormalization)                                                \n","                                                                 \n"," activation_43 (Activation)  (None, 32, 32, 512)       0         \n","                                                                 \n"," max_pooling2d_11 (MaxPoolin  (None, 16, 16, 512)      0         \n"," g2D)                                                            \n","                                                                 \n"," conv2d_46 (Conv2D)          (None, 16, 16, 1024)      4719616   \n","                                                                 \n"," batch_normalization_44 (Bat  (None, 16, 16, 1024)     4096      \n"," chNormalization)                                                \n","                                                                 \n"," activation_44 (Activation)  (None, 16, 16, 1024)      0         \n","                                                                 \n"," conv2d_47 (Conv2D)          (None, 16, 16, 1024)      9438208   \n","                                                                 \n"," batch_normalization_45 (Bat  (None, 16, 16, 1024)     4096      \n"," chNormalization)                                                \n","                                                                 \n"," activation_45 (Activation)  (None, 16, 16, 1024)      0         \n","                                                                 \n"," conv2d_transpose_8 (Conv2DT  (None, 32, 32, 512)      2097664   \n"," ranspose)                                                       \n","                                                                 \n"," conv2d_48 (Conv2D)          (None, 32, 32, 512)       2359808   \n","                                                                 \n"," batch_normalization_46 (Bat  (None, 32, 32, 512)      2048      \n"," chNormalization)                                                \n","                                                                 \n"," activation_46 (Activation)  (None, 32, 32, 512)       0         \n","                                                                 \n"," conv2d_49 (Conv2D)          (None, 32, 32, 512)       2359808   \n","                                                                 \n"," batch_normalization_47 (Bat  (None, 32, 32, 512)      2048      \n"," chNormalization)                                                \n","                                                                 \n"," activation_47 (Activation)  (None, 32, 32, 512)       0         \n","                                                                 \n"," conv2d_transpose_9 (Conv2DT  (None, 64, 64, 256)      524544    \n"," ranspose)                                                       \n","                                                                 \n"," conv2d_50 (Conv2D)          (None, 64, 64, 256)       590080    \n","                                                                 \n"," batch_normalization_48 (Bat  (None, 64, 64, 256)      1024      \n"," chNormalization)                                                \n","                                                                 \n"," activation_48 (Activation)  (None, 64, 64, 256)       0         \n","                                                                 \n"," conv2d_51 (Conv2D)          (None, 64, 64, 256)       590080    \n","                                                                 \n"," batch_normalization_49 (Bat  (None, 64, 64, 256)      1024      \n"," chNormalization)                                                \n","                                                                 \n"," activation_49 (Activation)  (None, 64, 64, 256)       0         \n","                                                                 \n"," conv2d_transpose_10 (Conv2D  (None, 128, 128, 128)    131200    \n"," Transpose)                                                      \n","                                                                 \n"," conv2d_52 (Conv2D)          (None, 128, 128, 128)     147584    \n","                                                                 \n"," batch_normalization_50 (Bat  (None, 128, 128, 128)    512       \n"," chNormalization)                                                \n","                                                                 \n"," activation_50 (Activation)  (None, 128, 128, 128)     0         \n","                                                                 \n"," conv2d_53 (Conv2D)          (None, 128, 128, 128)     147584    \n","                                                                 \n"," batch_normalization_51 (Bat  (None, 128, 128, 128)    512       \n"," chNormalization)                                                \n","                                                                 \n"," activation_51 (Activation)  (None, 128, 128, 128)     0         \n","                                                                 \n"," conv2d_transpose_11 (Conv2D  (None, 256, 256, 64)     32832     \n"," Transpose)                                                      \n","                                                                 \n"," conv2d_54 (Conv2D)          (None, 256, 256, 64)      36928     \n","                                                                 \n"," batch_normalization_52 (Bat  (None, 256, 256, 64)     256       \n"," chNormalization)                                                \n","                                                                 \n"," activation_52 (Activation)  (None, 256, 256, 64)      0         \n","                                                                 \n"," conv2d_55 (Conv2D)          (None, 256, 256, 64)      36928     \n","                                                                 \n"," batch_normalization_53 (Bat  (None, 256, 256, 64)     256       \n"," chNormalization)                                                \n","                                                                 \n"," activation_53 (Activation)  (None, 256, 256, 64)      0         \n","                                                                 \n"," conv2d_56 (Conv2D)          (None, 256, 256, 3)       1731      \n","                                                                 \n","=================================================================\n","Total params: 27,923,523\n","Trainable params: 27,911,747\n","Non-trainable params: 11,776\n","_________________________________________________________________\n","None\n"]}],"source":["autoencoder_model=build_autoencoder(img.shape)\n","autoencoder_model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])\n","print(autoencoder_model.summary())"]},{"cell_type":"markdown","source":["Training the model"],"metadata":{"id":"1fE-d8rORrWG"}},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":795545,"status":"ok","timestamp":1665077444739,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"},"user_tz":-360},"id":"JFIHMYmia5gM","outputId":"e6c2d397-e7ae-42a8-95f4-56c15ff9aa21"},"outputs":[{"output_type":"stream","name":"stdout","text":["Epoch 1/100\n","5/5 [==============================] - 10s 2s/step - loss: 0.1153 - accuracy: 0.3367\n","Epoch 2/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0521 - accuracy: 0.4077\n","Epoch 3/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0463 - accuracy: 0.4743\n","Epoch 4/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0433 - accuracy: 0.5075\n","Epoch 5/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0386 - accuracy: 0.4701\n","Epoch 6/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0334 - accuracy: 0.5926\n","Epoch 7/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0278 - accuracy: 0.5803\n","Epoch 8/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0228 - accuracy: 0.4892\n","Epoch 9/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0200 - accuracy: 0.3135\n","Epoch 10/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0184 - accuracy: 0.1961\n","Epoch 11/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0183 - accuracy: 0.1955\n","Epoch 12/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0166 - accuracy: 0.1857\n","Epoch 13/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0168 - accuracy: 0.1917\n","Epoch 14/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0163 - accuracy: 0.2581\n","Epoch 15/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0159 - accuracy: 0.2951\n","Epoch 16/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0151 - accuracy: 0.2791\n","Epoch 17/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0152 - accuracy: 0.2650\n","Epoch 18/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0140 - accuracy: 0.2953\n","Epoch 19/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0139 - accuracy: 0.2473\n","Epoch 20/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0138 - accuracy: 0.2908\n","Epoch 21/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0134 - accuracy: 0.3409\n","Epoch 22/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0126 - accuracy: 0.2743\n","Epoch 23/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0127 - accuracy: 0.2977\n","Epoch 24/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0119 - accuracy: 0.3300\n","Epoch 25/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0120 - accuracy: 0.3464\n","Epoch 26/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0114 - accuracy: 0.2714\n","Epoch 27/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0106 - accuracy: 0.3134\n","Epoch 28/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0109 - accuracy: 0.2841\n","Epoch 29/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0100 - accuracy: 0.3246\n","Epoch 30/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0099 - accuracy: 0.3089\n","Epoch 31/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0101 - accuracy: 0.3590\n","Epoch 32/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0098 - accuracy: 0.3769\n","Epoch 33/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0100 - accuracy: 0.3201\n","Epoch 34/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0094 - accuracy: 0.2948\n","Epoch 35/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0095 - accuracy: 0.3137\n","Epoch 36/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0091 - accuracy: 0.3279\n","Epoch 37/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0081 - accuracy: 0.3201\n","Epoch 38/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0087 - accuracy: 0.3541\n","Epoch 39/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0083 - accuracy: 0.2969\n","Epoch 40/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0086 - accuracy: 0.3224\n","Epoch 41/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0084 - accuracy: 0.2757\n","Epoch 42/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0080 - accuracy: 0.3190\n","Epoch 43/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0085 - accuracy: 0.2597\n","Epoch 44/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0079 - accuracy: 0.3194\n","Epoch 45/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0080 - accuracy: 0.2595\n","Epoch 46/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0072 - accuracy: 0.2973\n","Epoch 47/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0073 - accuracy: 0.2103\n","Epoch 48/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0081 - accuracy: 0.3382\n","Epoch 49/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0070 - accuracy: 0.2221\n","Epoch 50/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0078 - accuracy: 0.2408\n","Epoch 51/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0069 - accuracy: 0.2939\n","Epoch 52/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0069 - accuracy: 0.2927\n","Epoch 53/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0069 - accuracy: 0.3242\n","Epoch 54/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0064 - accuracy: 0.3131\n","Epoch 55/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0060 - accuracy: 0.2860\n","Epoch 56/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0061 - accuracy: 0.3169\n","Epoch 57/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0060 - accuracy: 0.2900\n","Epoch 58/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0072 - accuracy: 0.3265\n","Epoch 59/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0063 - accuracy: 0.2424\n","Epoch 60/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0062 - accuracy: 0.3342\n","Epoch 61/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0061 - accuracy: 0.3326\n","Epoch 62/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0068 - accuracy: 0.2700\n","Epoch 63/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0057 - accuracy: 0.3242\n","Epoch 64/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0056 - accuracy: 0.3117\n","Epoch 65/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0061 - accuracy: 0.3305\n","Epoch 66/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0054 - accuracy: 0.3437\n","Epoch 67/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0053 - accuracy: 0.3460\n","Epoch 68/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0059 - accuracy: 0.3171\n","Epoch 69/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0057 - accuracy: 0.2932\n","Epoch 70/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0052 - accuracy: 0.2863\n","Epoch 71/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0052 - accuracy: 0.3474\n","Epoch 72/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0049 - accuracy: 0.3808\n","Epoch 73/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0052 - accuracy: 0.3344\n","Epoch 74/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0050 - accuracy: 0.3921\n","Epoch 75/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0047 - accuracy: 0.3719\n","Epoch 76/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0047 - accuracy: 0.3280\n","Epoch 77/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0046 - accuracy: 0.3345\n","Epoch 78/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0051 - accuracy: 0.3482\n","Epoch 79/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0047 - accuracy: 0.3312\n","Epoch 80/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0051 - accuracy: 0.3730\n","Epoch 81/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0055 - accuracy: 0.3062\n","Epoch 82/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0045 - accuracy: 0.3430\n","Epoch 83/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0053 - accuracy: 0.3659\n","Epoch 84/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0047 - accuracy: 0.3370\n","Epoch 85/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0044 - accuracy: 0.3559\n","Epoch 86/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0047 - accuracy: 0.3690\n","Epoch 87/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0043 - accuracy: 0.3886\n","Epoch 88/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0049 - accuracy: 0.3741\n","Epoch 89/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0047 - accuracy: 0.3133\n","Epoch 90/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0045 - accuracy: 0.3851\n","Epoch 91/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0048 - accuracy: 0.3691\n","Epoch 92/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0047 - accuracy: 0.2825\n","Epoch 93/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0042 - accuracy: 0.3531\n","Epoch 94/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0043 - accuracy: 0.4030\n","Epoch 95/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0042 - accuracy: 0.3089\n","Epoch 96/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0046 - accuracy: 0.3196\n","Epoch 97/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0048 - accuracy: 0.4001\n","Epoch 98/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0044 - accuracy: 0.3706\n","Epoch 99/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0040 - accuracy: 0.3474\n","Epoch 100/100\n","5/5 [==============================] - 8s 2s/step - loss: 0.0041 - accuracy: 0.4080\n"]}],"source":["def create_dir(path):\n","    \"\"\" Create a directory. \"\"\"\n","    if not os.path.exists(path):\n","        os.makedirs(path)\n","\n","\n","create_dir(\"files\")   \n","csv_path = os.path.join(\"files\", \"encoder_data.csv\")     \n","callbacks = [\n","         \n","        CSVLogger(csv_path)\n","    ]\n","\n","history = autoencoder_model.fit(img_array , img_array ,\n","        epochs=100, verbose=1,callbacks=callbacks)\n","\n","\n","autoencoder_model.save('autoencoder_for_skull_images.h5') "]},{"cell_type":"markdown","source":["Checking the model"],"metadata":{"id":"Gz5Q_2rFRxD6"}},{"cell_type":"code","execution_count":null,"metadata":{"executionInfo":{"elapsed":1899,"status":"ok","timestamp":1665077684599,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"},"user_tz":-360},"id":"XJOkbbpta8Lf","colab":{"base_uri":"https://localhost:8080/","height":216},"outputId":"d4c4ab77-a0c3-48bf-97fc-9d1a280b768a"},"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 432x288 with 2 Axes>"],"image/png":"iVBORw0KGgoAAAANSUhEUgAAAXcAAADHCAYAAADifRM/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOy9e7BnyZbX9Vl779/r/M676lSdOvXo6u7br3uZuTMMwzBI6CgiQkiMQTgjCgqiDGiIjxDkoTxVGAxUDBBkNHgICkMMgijIMwSFGXFUvDF3Zrpv3zu3u6uqq+rUeT9+r/1I/8i98rd2nn2qqvt2U+feOavi1O/32zt37tyZub9r5XetzBTnHJdyKZdyKZfyjSXJiy7ApVzKpVzKpXzycgnul3Ipl3Ip34ByCe6XcimXcinfgHIJ7pdyKZdyKd+Acgnul3Ipl3Ip34ByCe6XcimXcinfgHIJ7hdMROS3ish/+0mnfY68nIh85pPI61Iu5RtVRORvi8i/9qLL8TxyCe6fsojIrxKRHxORkYg8EpE/IiKr56V3zv0e59xzdZ6PkvZSfnqLiLwnImMROan74Z8QkcUXXa5YPk0jQ0Tu1vlnn0b+F00uwf1TFBH594DfB/xGYAX4OcBLwN8QkW5L+p8Wne5SXpj8EufcIvAtwLcCv+UFl+cjy+U78vxyCe6fkojIMvC7gF/vnPurzrncOfce8L3AXeBXiMjvFJEfEpE/LSJHwK+qj/1pk8+/LCLvi8iuiPy22gL7J+tzIa2xSn6liHwgIjsi8h+YfH62iPyIiByIyEMR+UNtCuZSvvHFOfcI+Gt4kEdEfo6I/HDdN74gIt+laUVkXUT+uIh8KCL7IvIXzblfIyJfFpE9EflLIrJlzjkR+XUi8m6d738lIlKf+4yI/B0ROaz76Q/Wx//3+vIv1COMf15EvktE7ovIbxKRR8Afr0fDf9c+k7X4RWQgIv9Z/d4cisjfFZEBoPkf1Pl/Z53+V4vIT9bP99dE5CWT7y8QkbfrfP4QIJ9MK3z6cgnun578XKAP/I/2oHPuBPgrwC+oD3038EPAKvDf27Qi8lngDwO/HLiBt/5vPuO+Pw94A/j5wG8Xkbfq4yXw7wJXge+sz/8bH+O5LuXrXETkFvCLgC+LyE3gLwP/MbAO/Abgz4vIRp38TwELwOeAa8B/UefxTwC/F2+s3ADeB/5sdKt/Bvh24JvrdL+wPv4fAX8dWANuAX8QwDn3j9bnP++cW3TO/WD9e7Mu20vA9z3HI/5+4Nvw7+A68O8DFaD5r9b5/4iIfDfwW4FfCmwA/wfwZ+pnvIp/f/9D/HvzFeAfeY77Xwi5BPdPT64CO865ouXcw/o8wI845/6ic65yzo2jdP8c8D875/6uc24G/HbgWYsB/S7n3Ng59wXgC8DnAZxz/49z7v90zhX1COKPAv/Yx3u0S/k6lb8oIsfAPWAb+B3ArwD+inPur9R98G8A/zfwi0XkBl4J/Drn3H49+vw7dV6/HPhjzrn/1zk3xVM83ykid839vt85d+Cc+wD436hHCkCOB+ot59zEOdewwlukAn6Hc27a8o40REQS4FcD/7Zz7oFzrnTO/XBdxjb5dcDvdc79ZP2u/h7gW2rr/RcDP+6c+yHnXA78AeDRM8p6YeQS3D892QGunsMR3qjPg3/RzpMte945NwJ2n3Ff2/lGwCKAiLwuIv9L7Uw7wnfiq20ZXMo3rPyzzrkl4LuAN/Ht/xLwPTV1ciAiB/jR3w3gNrDnnNtvyWsLb60DYUS6S3Nk2doX8Za0AP+XiPy4iPzqZ5T7iXNu8pzPeBU/Yv7Kc6Z/CfgvzbPv1WW7ydn3z/H09/VCySW4f3ryI8AUP9wLUkco/CLgb9WHnmaJP8QPW/XaAXDlY5bnjwBvA68555bxQ9GvG/7wUj45qa3vP4GnL+4Bf8o5t2r+hs6576/PrZ8T3fUhHhgBEJEhvm8+eI77P3LO/Rrn3Bbwa4E//IwImfgdOcVTRXrvTXNuB5gArz5HPuCf8ddGzz9wzv0w/v27be4j9vdFl0tw/5TEOXeId6j+QRH5p0WkUw9Z/xxwH89lPkt+CPglIvJza+fn7+TjA/IScASciMibwL/+MfO5lG8M+QN4v88P4/vYLxSRVET6tRPzlnPuIfC/4sF3re7Dylv/GeBfEZFvEZEefiT492vK76kiIt9T8/4A+3jQrerfj4FXnpHFF4DP1ffu498LAJxzFfDHgP9cRLbqZ/rOuoxP6vvY/P9r4LeIyOfqsq2IyPfU5/5yfZ9fWo/A/y08//91IZfg/imKc+4/xVvIvx8PrH8fbyn8/KdwgPb6Hwd+Pd5R9RA4wXOlz7y2RX4D8C8Cx8B/A/zg05NfyjeyOOeeAP8dHrDUqfgE3z9/I3Ns+JfwHPnb+L7379TX/03gtwF/Ht83XwV+2XPe/tuBvy8iJ8BfwvPjP1Wf+53An6xpku89p+xfAn438DeBd4GYs/8NwI8BP4qnWX4fkNS05n8C/L06/5/jnPsL9fk/W9OVX8SPrHHO7QDfA3w/nnJ6Dfh7z/mML1zkcrOOrx+pKZ0DPLXy1Rddnku5lEu5uHJpuV9wEZFfIiILNaf5+/EWyXsvtlSXcimXctHlUwP3mmd+p57k8Js/rfv8NJDvxjuvPsQPC3+ZuxxuvTC57NeX8vUinwotIyIp8CW8w+Y+nvv6F5xzP/GJ3+xSLuUfklz260v5epJPy3L/2cCXnXM/VU+++bN4C/RSLuXrWS779aV83cinBe43aQb73+fZ0+Yv5VIuulz260v5upEXtsKaiHwf83Uivi1Jnk/PiAhPo5Kedf6jpvu05dMqh+b7PPm3pbHHzjsPtOb9tVz7LPk411ZVhXPuH9qErbhva5n9OfBTFRyXc8jaRaSebRQ3sbQc+2krDud8dbiqau1Inxa4P6A5k+sW0cw159wPAD8AfkW3wWBAkiQBkPS7/sXH9IWxL7sFM/tCmXs2rqlfetI0paqqcE7Pl2XZOGcBJb6/PZ+mKUmSUFVV+EuSBBEJedl7VVXVKJNVdPosRVGQpmm4rg24bX76XDZPvcbez14T31PrVMtTlmVohxjAAZIkoSzLRhltu7S1iYhQlmUonz1uy6H1aX+rtNWv3s85x8nJyZn7fkx5Zr+GZt9Os9R1Oh0SEQSp8Vya8P4srG+ck/D/edhnz8XHwv2iMjwPjoqAc3q56feA1GAjgKsvDunnN30uadxb83pKWeeiJXlGeWxeT8m/Nf1zPYFrNMCz8pifc7ioM+gxrUIHUDlc/XzT2ezcUnxa4P6jwGsi8jK+8/8y/ASac8U5F15yILzIFoRioFAgscClaTUPzc851wA1BZIYkGLQKsvyTN5aDr0+BuxYGWRZ1nimuFzxs8RgC9DpdEL6NE0piuJMHdlypmnaeI42YIyf24oCedPqnNdrXFathzbFYMsZK2CrcGKAt20bt4/tB3Gd2rZTBfcJyUfu1x5EBZAAiB7kHeIU7Jl/nkHidmkDGddyLj6mQOFM/uGcAcAz5dFrNH3UTjaP+LRTM/w5Byuu5Ufbs7VeKK61PHp7N9c2YXTQVhcWnxVw2+opVs7+nDSUa+PapzyvO9MZAJKgRyXS0DLvUq3yqYC7c64QkX8Tv2Z0il897sefdk0bwLQN7WNgbwNCCxoW0GJQ0+MxQJ43UojBVK3yGET1nIg0QLbtmS1QWuUT14MFPwuiKjG4xiMGBTkFbVUS8QjAWuE2X/2u6ayCi8ts09vzbfXQ1k7xKEJBOr6+rSz6qW0c942vRT5Ov257nV14jVvQTmobTuavujG3P4KZbbI/g5bWVpz/H1+goCbGfmy9yrmaSpFQ/nkehFyaeT39XKOM2kcb97blabubnCmXGI119ur5L8DXv0gN1q5Zh/YZpVn+kLBG/3gE8pHbTeal0mdsU35tciFmqIqI6/f7gc5ooxcssFrggqZFbK05oAGIsWUeA0tsRerxmMbQa631fh6lEN+nzYqN6Q4dLVjr3t7b5hP7KhTUbFnicqu0jYJiJdtGhZ1HLcXUydPO2fq1owR7D3t92zPYctoRja035xzj8ZiyLJ+fF/gEJc1S1+10zxovNSdwRpcHC1ABLbLw50ma6WkkebrUwBMA0zlcMEXrcxjcsOn1nGKbPfZc937ehB8zr6cCpnm2oIdk/tjMsbRxTPOJz0XYGY9IY/lIbaQKI7RRfW8dajivTibTCWVRtGZ7Ibes0pdSQcFa2va8FQt0FvTbqBj9bq3d8xpFRMiy7IwlqJZkDDTxPfTP0gP6bFpWC+yxxMB2Xj21nbfliamZOI80TRtWvV4TU19tFFVVVa2+ifje2i6aNh6B2DbXZ7f+C5vuPN4/VuwvXFz7Sz0f6DcBfz7qVvu1wbqGa+MfZ1o/gDWNofwZ4KYGaltgPRbhk9qM7hnHPp7Eo4fzjp0p6ke8x/yypk1cf5Ozx8LV5pyL2q1J45y9a+vbaZro7DnXaJjQB1x9YTyaaJELA+5tVqqlDWKLUMUCik2rYi3CmHLRzzaL35bF/rYjgfhcG8BbeiemdWyamMqw5+x3O1qw97Jls3laALbp2yinmII5T2HYcsWjoXh0EUtcd0+73iqO2IKPFUhbHdjRzwsVMzxX4J4X27z6zlvRFpjnL3/tIDSWchBjfeo9Qr0oS+BiSDL3ZA7Slg2w4BfDrH0eYxRHB03CZ4pEn3FGRN+fnn+s286UtZGT/RapXDsyMifEmToJqO4a51pM/WbhTANr2wZx9X+mnzhnP0Oic+UC9HwvlhO3cp6FboE/y7IGaMQvf1mWgeqAuUVvnZIWzCzQqPWaJElwjMaAe96ffYbY2myjOGKHoM1f09rRgF5j6yi2aG2+MaUUj16sJR3nYSN/rMLS/OLRQ1yf1lKPR0ttdWl/x89ryxI/s6W49PdFEocBg/q7bQ9/3oVz8TXhnP1z/pvmVZm0znyCX+/W/zZg0SjL/J5n723OxenacOZ5jzVOO5rAJqaEz5+XRr62gng4p3A/f0DnJNzaMa9XtM5arP/60vrT1LlrGl+a/tx2rhugyVBouZiXRXN6Rl1eGMs9ftlj6iXmUWEO4homaIGh0+kEEIlBM7Zw9VOBQe/TBsRtnHJb2S1N1AZi8YjBpmtTMpq+jWO35bOjhDikUj/bQi1tXejx2NegSjRWpPpp66Io5rsL2rrQ38qPx8q7LRJHf1sFrXnaaCZL58R1/iJlblV7qJEIciAyQj+SxTtPL+LBSdTErGkEF+qhvnc9qtf0ejMRm5e1xGs6xxqh9eOcZw23HrPW/JlndOFRNOJFP8Wcb+Roy2ou8/SJzVzrxPRXrQRLeIlx6Trj4JV5WCXhWFOhhToPZZPal3He856t69BWzigccbj4uhrYz1V6tVwYyz0G7RhYraM1HqK3xcfrS2+/twGpTRNTHbYs1lLU49AEo1hh2FFDbMVaP0L8Z8Va3dYabvMVWEC2MfFxuS19YevRWte2fLGysfnZelZAj6mwuH3tfe1zxJFDcZvFfcLeK5a4zC9W5lao/9WEq/i7NcoaVrJrSXvmvAGDM/O25hk4k75xNrqXvS624O3V9lybVXtemecyrxPnzvlseY5w7zNpGuqikf/8WVvaYV4xZy3r6FjzOqPW1NqObvj0dtMynV/X9lmaz9YuF8Zyj61H69yLAaLNIo655TZr2IKGSjwKsN9tOutAjQGjDUAsnRPfJ7aQzzt2XkggNEMn26zxtrpVifl9O4JQ0LdKJa5LS4nFitiWO54cFtdtrJhjBaVlbGujWLG1ndcyvHiRxnuozXHuq6lGpaaZG4Pt1xkLsGEqx5ZpS/5xNk8dNcT5Ny58yjE55zNSIOfG3IhNZZ6oMUx4ehHmV5octH6i0MZGnudY22fqUMwXrUjbP8+p89ZCO5rXYUC+bWR1jlwoy91aj9bytFa7BXB7nabVv9hiji3tGBQsAMZWqw7/rZWp51VicD2P7rDnY0Vl/yzFYoGuLTLGjk40XQystszx9fa8VX7qB7GjG62H+FlsPlqfasnbZ03TlCzLGs8RKwBLscX5xxFRVtra58JY7h8x+Rkr9Xmy1+siy71RBbXJ2RwFtJjiTxE582VOSEjQY2L+NX+Ff/VFIppe6RL9POdavc78nqeb5/R0DdN+TJ5VB7auTbo5cM8tdw/KzVHGMyUaCkl9H1VoYr49Sy6U5W6BBJoTaPS8gn4MoDF/bQFL07fFurdZzfGEGRtp0uacVCBTMGyjFGx5z+OCY6C3ii2e1m/T63mr/Cwww9yKbuPbrQWv+WpecQy+bZP4M65Hy5vb/Oy5+DrNzyr7tn7RNvqKn/FpYaQvUlxsZYcDTz827+/GVI2+Ng/NrWE7xUZMKsGhMX5nKa85sDidBFSH6InYO501nUUzwCHWhmyYnGKMZWeulZbvZ30Jzbvp9Ul4Wj9HgKfz3o26qOmV5wgzDNdp3YUaJnDhZ7Noa7f2Yce8Lf1zz/uMOzPIOE8uDLi3vbAxCLS99Jo2vu48oLeAVhRFA9zivNron/NCIOPp/vb+1qpuK19bvL1KmyKyESBa/ja/Q0xt6F9M7cT0k71vHMYZUyi2juL8Ysu5DcQ1jv28NG3KN6ZurFKLaZ+LZ703Qcofisf6FjyJjvk8OOdrDBUKr4GOOIMI9p4yBw4xRQt5RuUO35twrBdaC1PMb1eDpwfEyC1o+4v5Jt4D7O1xScKtmy1bhye6+Z3nk5U0q/gKsL4JkWe4Kc1NG++pTSIGgxuFtI2lhWqH6JCHpY4aI7FnuVMvELjHseYxQFiL2laqgo8Nh4yt5XgIr2DYNkEqBkblzWNuWfNto2ws99z2TG3fY0Wi91cFZDl/vUaPKRcfKzvLNytFYp8lTmOfOR7ltCnYthGG5mOf3z6XVc4iQp7njXpsG1m0UUqxEtK212NxO75wcRKQ0tWWrzXOA3w2ztXIECzrNnA2tzCf0jjmGhapzcNanDYTZyxdEeJUKC0SsLxhYc9/z/uEnvNlca4C6nh/Nw/fJMD9XC0hdZ6S1J/zY8ZG9+1d/3LOGcs9tsTrewQAVmVDM10beppj4X6NqJ65fjqTXyOfM8OP5unQDtpu87Y/dwASyYUBd8ulxrHKKhYYwHacsxN1LCUQ0zwxEFiJrb82xWAtWgUmvVc84UbvrbNcVRHZ++nzpmlKr9cL4Kj3TtOUhYUFOp0O3W43KDK14pMkIc9zAKbTKePxmNlsFuqyLMvGLNs2SiMG3zZwjNsjBv+Yi4/DPdtGJvFMX803ni1rlV7bsbiMVnleCAkW49y+biuaPJflHl/EHIRDKCRz61bmlrsz5/xHZO/H5n84PNcKHqclgLsgSGLYYAFhbl3PgTkJgxRXP58kQpokJGlCmqRIAonMfS7WOCjKglk+I58VuErVzVyZuLqMrqpwOKrK4VyFc3U/D6g4V3YOoPIA39QBTp/6jFKLLfc5wGufMwOQ85D4Wd0ytGnd12VenueVCwPuMH/RlTu10SGxNRpbsHEYXgzoCp4q8YggtozVerWx3SpWqVjeO8uyYI1a+kPztHnptZ1Oh8FgwNLSEktLS6yurrK0tBQcj4uLiwyHQ5aWlnDOsbCwQJIkzGazANz9fp88z8myjDzPOT095ejoiMPDQ0ajEQCTyYTDw0OOjo6YTCacnp5SlmWoNwXmTqfTCB11ztHtds/Um16bZdmZOrbzAc6r95iSqqqqsfKl1pcqLVWatqyWotI2tG1v075omVvpNVy42mYWe7JGF8OxugAkesyOpOZDdb2H5hGsVQknAoDN+Qq98qzljotPWfBWRZOQCCRJiohZSiNNSCUjSbxl3cm8r6yT9UizhCzN6Pa69Ls9VhaXWF5dZnV5lYXFBbJel06aQpqQIZQOinzGbDZlZ3+Xve1t7j1+zOnRKbOiAIRUhEpAKt/eRVlRVTllWVFVJVXlqCw153QSkQf+Kqk8wAPqwWxG7sQaT6vIIU6alv+8uuf1GJo8ajc7etOhUnPI1cwraJ9oMtM5cmHA3TraYooE5oDatjDUebMmoUn35HkewEbXUonT2t9tESdKf8TOQrWMVSHY2a8Khp1Oh06nw8LCArdu3WJ9fZ2lpSUWFxcZDAZMJhNGoxFFUVCWJQcHB4zH41CWLMsYjUYMBgNWVlYAyPOc8XiMc47ZbMbe3h7b29uMRiNOTk6oqorBYMDGxgZ37txhMplQFEUA+el0ymw24+TkhOl0ynQ6paoqptNpqG/1b2gd60hAy2XrMV6LJnbqWnDS+ut0Ot4yqykoFQV2S9nYUY0tj46aYi7+ooC7L5MBB8EvNdBkaw1N0mIFhrzsrzbQMZasuahmRlry4Cx+Bawx1rla0+Z7kghpkpJmCZ0so9Pt0usN6PQ6pOL7bLfTZdAf0uv7d6eTpvR7GcPeAquLC6ytrbK6dpXl1WWyTpc0TcBB6RxVMeN0PGU6GbE0XGJtuEKvv8z+/j6TYoarEkgcrhKKckaez5hNJkzzKfmsoCxyisLMUXEE2iZQjJVQJQ7nSioF+aBon9Km2o7abi1t5RPGI695OwsRqLe05Hn39vc6v4AXYlXIJEnccDg8swRtGwUT8+Txy66zHs/jtOHsZhp26KfX2A0kFNTbOOjY2o+5516vR7/fZ319natXr7K5ucnKygrdbjcA+P7+frCyRYTpdEqe5+E5syzzL0m3G/LXzU1EhOvXr7O7u8vBwQGTyYTJZMJsNmtYwmr1q4IpioJer0ev12MwGDRoFKV28jxnd3dXV1VkVm8MEDtN9ToF+jbKyzqc9RrLscczVm07az+w0VJaDhXrg7DP7JxjNBq9uFUh09RlWeY360hqGgEiq9lYiLUp5/tUfdjF6c8cbOSktAnnpbQ0js3xzDEBkbDRiL5nSQIkCVnil+Tod/ssLAzp9YcsLQ1ZWBiQBuUL3U7m27fKmU4K8tmU0jkoHZImdDsdhv0BC8NF0qRDmgpF5RBXUZQwnU2pypxZXuKqgqKoyMucFJA0JeskOKAsSmZ5waQ2WKaTMZPplGk+w5WeqnGVN6WLssJVFUVVUuYlVVVQmkl8lR2ti60PU6Et0Knt1mCBWiXWpHH7npW58nA11QSzfEpZtPftCwHuIuKWlpYCSMTx3G2OMguqluZos9Rix5wFo9jSV048Ph5HjbTxyHrvhYUFlpeXWV1d5caNG6ytrdHv90N+jx8/ZjabUVUV4/E4gLHGf6v1rNa+jjrKsgyUj5bFgl2/3w8jCBFhYWGBqqrCvfI8p9frBZDs9XoNuqPf79PpdIIi6Xa7jEYjjo+POT4+Zn9/n+PjYyaTSeP50zQNjlGrWGNHdJtVrdZ2PFpqo8Ksko65e6tkNL1ec3Jy8mLBXXdiUuv5XEuthb4xQ/hzJYC1/3I27PFsuvBO1DcNgGRNeoEEdWImpKkEQ6PT6dHvDegv9FlZWmF9bYnBcMBCt+tBZzYjLxyl82BcFAXjyYjRaMx4OqHIi1qZ+/bK0oxOlqG8fKUUE3NF5xCvUNKMJEvpZN5Q6XW7dT+uq8s5KiryWcFsNmM8m1HmJaUroYKqKimdo8wLpvmM6XRMPp2QFyVlVeIq55WPc7Xj1fPz8tRG0PabO3WhfeSFbd7nEP9YgiS+reajD78T04Ve8je2gvM8D1ZnDKBKCeh6MhYAYnrFDtMVMOziX3F6S9koxRBb/frbUjIKkL1ej42NDW7dusXm5mawrkejEfv7+4Hn1mesLUo6nU7Dkr1y5QqDwYAnT55wfHzMbDZjNBqFEYdzjul0iojQ7/dJkoTxeMzp6WlwuuozKPUzGAzI8zyMVsbjMd1ul8XFRWAOxGVZBoWztLREv99neXmZPM/DKOPx48eMRiOm02mj3FbJWpDQe2odW7oqdmDH/SKetBWnjwE/7k8XYobqHLGf+ULboXv4lKcjgXfo2V8mh4aVPne2zp2B8/8scxTUUOL7RpZ26PX7DAYLLC0vsrxY+4gWllkYdukO+lBWTPIJk/GM0WRaOz9LZrmn/ybjCePJhOk0pywLqrKkrOp2D9SPAUM9loCQkKY1d9/JaiOkQ7fbYdLp0u12yNIOWZp66rPbJ1vMAN/H8qKkqkrysgQHlasoq4rJZMrh8T7HxyecHJ/iphMqV5FUrl7nxerWRsFazXLrZLUK+ow1/7zArm0zz6n9xi1yIcDdgncc9wwEMLfcdpvj1OYRn7ehc8rb2/QWmBSA7ShCr7UKRcuaZRlra2tcv36dV199leXlZcqy5OTkhMlkwng8DhZ0URSBfuh0OgyHw+BULcuSyWQS6JqDg4MAoMvLy6yvr1OWJXt7e4B3fl6/fp2bN2+SZRmHh4fs7++zs7PTUCQKgP1+H+ccvV4vOCtjh6iem0wmoR0mkwmDwYBbt25x5coV1tbW2N7e5tGjR6F8qgwtEFvQj+kq68+wnH6ssGP/hgV3rQM70clGW7WV6UWLxfnGQQXxeEJRy7mzeSp9MA8HjDHAZ3NeHnpsPrIQIEmErJPR7/VZW11n8/p1NtavsLayRNbrkiQ+y7KsmMxyZtMpJ6MR49GEyXREXhTeei5mTKcz8nxKPsuZFQVVWVDW1IhzdfCj4aHCY1t+v0zIspTKVVR12xZVSacsyMsu3aygk3VweMzopYk3dBRPqoqi9Lw6ifNROZVjNrvG0ckRHz7eZntnm+PTEc7liHNUpg2Uiw8Dq7b2RfnwpgWvTeTOJAwXtHQMveA8Xv7p/fpCgDtw5qWNJ8RYMAAC+EKTk7VUjqVj7LWxZan5xSBlJwfFVJCCYqfTYXNzk5deeolr164Fx+h4PA7OytlsRp7nFEXBYDDgypUrpGkanJtPnjxhf38/KAO1OJWmKYqCPM958uRJiJIREWazGY8fP2YymbC5uUmWZQGAT09PefDgAV/96leD8rl58yaDwYDxeMzy8jJpmgbL//j4uNEGi4uLVFXFZDIhyzJmsxlpmrK6ukq322UwGCAiPHr0KCgr5dxtnHybRR5TNNqGQBhd6LE2OsymtxO6bDtbn8pFoB5V2slRmL/A0Yvcdi6+3Fi7oU7t9QE36nwb1nwzR3X8igi9bsbSyjK3N1/itVfusrV1A0kgL5e//0gAACAASURBVGdMplMmkylFnjOZFEzyKePxhPFkzGw2Ic9z8rygKEqKIve0YlFRlg7qKJWaXDBRJp4PdzXt4HWRIK4OtXRQVUJZVSTa7qoAmIVo00SEWScjKzNSl5LWoZVJ4qNvoKAqoUpKBlmX4XDIzc0t7t64y1cfvc9X3v+A7e2HzCYzqOZ01zwixtTaXB/Nm8i0g1haKabWmxV/tk+06mJnJkY9vV9fGHCPJV5JMXaixi9sHEFjh/TWMrWgr39qYVqQUlEqx95fLfbhcMjt27d55ZVXWF9fp9PphAiPo6OjMyGRi4uLLC4uIiJBAdy/f5+DgwPAc979fp9ut8t0OuX09JTJZBL4bGuRWqv65OSEhw8f0uv1WF5e5u7du2xsbDAcDtne3ubg4IDDw8Ng6SsoK0c/m83odrs454Ly0PIuLy9TFAX7+/sURRGon7W1tUAnPXjwoMHDxzRWzKtD+xo/dpTRRttYZRC3tw2JtKOui2S1w3lUqzlSv81NflfOpGsa5u0WvT9zNjba0jghZ/HUB+Lrtt/NWFtb5/XX3uKzb7zJ1StLTGc5O7v7HB4dMJnMKAofbjgrZkzznHzqo1XyvKAoixD1VRQlVVlR1cCOq0cG1CMyNJLFUTkBqhokazpCLdu6AsV5x2hFSSUJVVJSlUKZlEjhQTyZTEklQYCq40gSn1euDte8QKgYd7osLAzIOl22bt7gxp1Ntra2+Adf+AL3H9zj5GREWVn/hzaRPxZPNG2H27kGaLTbU7A5RLI6IuAXsxnI0+XCgLuNF28DcH1ZrbQBQxyhYUEldoLa49Zy12PWktf8RIRut8vq6ip3797l1VdfZTAYeIdRzXvv7++foRyGwyGz2Yzd3V3yPOf4+DiA9dWrVz0nWYcmHh4e1paPp01WVlYagDWdThmNRmd8EScnJ+H+V69eZXV1lSzLWF5eDukmkwmrq6uBd9/e3q55TO+41dGGUkbD4ZC1tTVWV1cbSrPb7dLv9xkOh/R6Pd59991AOWn9abu0/dY2j3n1NiC3yqxtUplV2JaiscdfqDRGKM9K235gzg5I7ejjzDty9rq5U6+ZCdihfrAxRUhS78fZ2LjG5976LN/8xhssLy9ycHLC40cP2d7Z5eh0VFMqJThHWTmKMvdOy6qkrHTikHeM6uQiSUBcQlKBpEAKzlWUVf1OVv4CIalZIp2glCCJkIjn3SVJvSWfJOYxpb6nD0TIyxnTWR1FUzlS8fZ9XhScTsZMx2OKqiRNMhbGE6rCkXUTXr55m5/1Ld/G5rVNfvQL/x8/8eM/xsH+IUVV1sxJPQPWsOrzGm/nw239N2gbbT9z2fzYU9y37bc5IxcC3C0Foty6BS7LnwKNaBZr4WkUiFrgMbUTz5a01ruetxOPLOgoDZMkCaurq7z22mu8/vrrLC0tcXR0xMnJCUdHR4FCUb5cJwDZGHrl3kUkcO42EkZpjcFgwGuvvcba2hpACIU8OTnh7bffZnd3FxFhaWmJtbU18jxnMpmEyJbZbEa/3w9culI9qjAU7NUBq+Uaj8ccHx/z+PHjoBzW19dZX18PPgsdZayvr/Pmm29SliVf+cpXQjtYusT6MaxVbrl3a2nHNJjWe5uijh24dtet2O/ywsTwqedRq896XwPtggJ7/TsM/5sULtBYzn0OMJg7zbnkpG6Lbq/L9Y1rfMu3fCuff/1N+sMOh0enPH78mIc7T9jfP/TRXEVZg6/XDq5yfsKQcyT1yEDELxnmJMElDiEFmU84y9IOaZYxK0qKYuqt/uk8giZNNdoqIanpojSbLx0i4me3+pG0D7tUGqQsKvKk8FFKDqosQRzMyoKiDvUt6klQ08mUvJhSSkWv1+furbu8+dprDBcXSQS+8IV/wOHhCY4y1K4z9Vrb8I32aNI4purP+d42oouPBSrrObvz1wTuIvIecAyUQOGc+1kisg78IHAXeA/4Xufc/tPysVqsjYfV7/ZltU5Ta9UrhQE0aBlLwVjnqaVttNPAfE0Xq1iqqmJ1dZWtrS02NzdxznF0dIRzjn6/T6/X4+TkhAcPHrCzsxMA3VtCniaZTqcMBoMQbaLce57ngcfu9XpMp1PW1tZ4/fXXwyQfLceVK1fo9/u8/fbbdLtdNjY2ghN3Mpmwvb3N3t4e4/GY0WgU+PKFhYUwq3Q4HLKyshK4dy2LKsijo6Mw03V/f5979+6xvr7OtWvX2NjYCE7uXq/H9evXqaqK4+NjPvzww1bFa2cR29GWlkcnS7W1t+0jNhLKTpiKY++t4/3jyCfVt31mzZ8B4O1oP+Juz9MCgZ0wnKtIk3oJvLBBBxc78LRgzgNTkiZ0uz2urK3ymdc+w5uv3qU/6HNwfMDOkx2ePNnh4OCE8WhCXsyCQpAkJRFjndfcflqBS8AHo5feynbQwdFJO/SHfVYWl1hbWaF0jvHpiP3jQw4PD5lOZ5AkdNMOktURYpWPrNFxW4IfHSRpSlIriyRJ5koUr2yKSqAUKuct/qqqKCpwzvel0lWUrqQ89riQJhkZjo0bm6yvrvEz3vgcB7vbvDN9n+l4RFXXGaGvGaI9bitD5SDa740aD3mYdjN9OLSRjhRi3v4fgkP1H3fO7Zjfvxn4W8657xeR31z//k3PysS+2G1WnAXjtkkyMAeA+DoFxXiGpAKPgoAuH2CtTjuSWFhYYGtrixs3blCWJaPRKFjGaZry3nvv8c4773B4eEiSJKysrHD9+nVWVlZCyGFRFBwfH3N4eIhzjuPj46AcdN0YpT2uXLnSCN1UPl+BrNPpsLKywo0bN1hcXGQ6nbK/v8/GxgaDwSDw7XZt+16vx3g8DpEz2omKoghUi0bnLC0tsbu7Gxy9jx49Chz/lStXWF1dDQB/7do13nrrLY6Pjzk9PW1Yy9YCjy1vVaj6nNoe5/UPbXO7omfM1Wt6S/V8TPlE+vYcUGsbr9VpGn0/p8z6giuI1uhcb5tnX/4m6Gh2zc2Z6vIkQjfLWFle5MbWLW5cuwYJPDnY5eDwkP2DI07HE4piRuV8KKHUlnkqOjmLcJy0NsCqhLL00S2+DH79mH6vx9r6Olsb17m2sUavM2Q2HbF9uM/9D+5zeHJMWUGv12GhPyTNMoqqYDIeMS1nzCYzinxC5Txlk2YJiSRhfRvP0/tvVeUo8fSROA+0nSyl6nUQ8cZF5RxlUXByMmI7fYxQcTKZsLZyhdWVFV5/87PsH5/w8MOCvMgb9EzDNm8Mv+bt7Oo2MvZ9ZNCadtMmDTRc/b4402ca1Nr58mnQMt8NfFf9/U8Cf5vneAHsOiVxhIrlUNWyjgHAWnkwB3oFDWvFKfViAcA64vR6K/1+n6tXr7K8vIzy7p1OJ1jGX/3qV7l//z7Hx8f0+31u3rzJ5uZmmIk6nU6D9X90dBScmEqJJEnC+vp6sDo3NzdZWFgI8e06O1Rnl6qle+vWLYbDYQDIxcVFrly5Qp7nXLlyhUePHnF4eEhVVUF5jcfjQM/0+/1AYylYaj7dbpeFhYUw+3V/f5/d3d0Qrql1q89448YN7ty5wzvvvNOIWmnzb2j9a7k0fbxgmAXwOCxVlbbl5DXPp+2c9TXIx+rbc2lD848q0rz8zNoy9g7tRIA9JyJkacrCcMDGxjWuXr2CE3j4eIfZOOd0NGI0GpHnM1zprXWX1v6PtKZIDJeMeBAXoKxS0rQKiiBLfHz64sKQK1eusn51jcXlRfrdBcT1yfoZVZ4zOOiRlwUL/QWGi4v0+12yRBiPZ8xmE7YP9jjZP2Ka5369nCRFww6dg0Q1nTjTRyoSV4cupxn0BqRpxnQ2oyhmwTF8eHREWTlG4zGHK0esra9z7coGr9x5mZOjUw6PjnCuJKwro5z5eW0VGucpyc5Y49GQztJrgRZ6dp/+WsHdAX9dvIr6o865HwCuO+ce1ucfAdefN7N48lEMtDFXq2LpGU2vL3y81Z3y9ZZ20XNqEVu6Ru+lYKd0gFrn0+mUDz74gPv37yMirKys8PnPf5719XXG4zEffvhhmIGqtITy2BrbrqGFg8EgALCID3XUcmnkgQLc6ekpvV6P9fX1xmbUvV4vRLFsbW2xvr7OwcEBx8fH5HnemP2qoxRdWqDX63F6eho4eZjHmWu77O3tMRqNePLkCWVZcvXqVfI85+rVq3S7XV555RUeP37MkydPGpazbRsL3jZkUY/bOQhx+6hSsRRebO3HS0N/TMv9k+vbDXpx/umshQYtVhtNR1ud5sz2bhEezCM3XCsYuHCZ56wXFvpsXLvJ9a3rdFJhZ2+fsijIZ46imFKVfjGuJBUy10WSAsFTIpmOhmv6QZL6vTGKLE1qX0+nQ6/jKcHl5QWyeqJi6Upc6ZAKsk7GoN8lKTO6gx4LgwGDhT7DXp9q1VHkU/qDAY/TLsenI/KyxKFLCRRo4I0E/r3CIVQlpAKJ87H73U6HvPKzsceThMl0RJGXIbpnPJkxGk0YzWZsXdvk1dsvc3h0xLtf+TKnI7+Wk6upoTPkeMOCP+dYI/28E9hkjX4rxjBSov8Z8rWC+89zzj0QkWvA3xCRt+1J55wTaSekROT7gO8LBamtQAu8dkJL29TyeKZjzLHbSUiW6omjKdo4f02na7GoU3EwGHB6ehpCFXd2djg+PiZJEjY3N9nc3GRxcZHT01P29vYaIYdKrSjnrRNwul0fa6tx53mec3JywsLCQpjaryMbpWPsImi6vID907pTxZFlGaenp6EM6mwdjUahPjQeX527umTC+vo6WZYxGAxYXFwMDuQnT56EVShFhI2NDdbW1rh7924I49SZxLbeTf9ojJb0eeK0Fui1zuwozraZtfytYnhRfVuaPIrhXzWf+hpajlna3Fh+bc63kAd28C9nzun3Gi3o9bpsbt3irdc+AxTsHR4yHo/qtVe8ckgFbx07PAXiOvWoOCVNMiSZc88+osWvpaP+m27XzyTNOindTo9Ox/PkVVEwnsJ0WuKqktPxmDwvyEtAOfGqwlVQAN2kS9JLWFn2oN3v9RjNJuR5wXQ2I8kTKnGgoIv2sbp/FYLrQip+dJrRIRVvAM6mGVVZ9/1ZwWw6Yzad+JBJSbl5dYM3Xn2VvYM9itkjpnmJzviNWZLzcLxxzDaiAfbnkufsz18TuDvnHtSf2yLyF4CfDTwWkRvOuYcicgPYPufaHwB+AKDtJbHWnYq1wOPQSGtxt73YaoHaWZuqAHQUYFd11DL0ej1u3rzJK6+8EiJf9vb2ODw8ZHt7O8R937x5k7t377K4uBgcpNvb20yn0+BM1YW4er0ew+Gw4UhUisQuamaXPFbKKsuyMJLQFSRFfNSL5mNn2FqLVpcmKIoihFPaerETpA4ODsJsVV0np6oq+v0+N27c4MGDBzx8+JCdnR1EhA8//JDRaMS1a9fY3Nzk4cOHbG9vN2gRq2xtmGobpRZb+G3tqb+tgreO2Nh/8lHkk+rbaZqaFcCecj/7RZrHnPlfaCqLpt6oo9pb0L8xEqhH+lmWsrF+hW/+3FtsLK/zzvtfYXd3n1k+xVUVknjLPMtSstR5GkaSekmCtA6hTT0NIz4qRdKETprVTs6EbpbRGwzodv1ospumVOKjaGbFjLyckUpCXlTMpt4YqJf4onJQlSXTfIYkQtWpEPFrwHe7PRaHdd9PpriKmgqqqMqaoqkcOP/pV230a71r/lAbiklCkvrPqsgpq4qyLCgr33c63YyFxQEbV9a4c+suJ0dHFEen3iipZxQ16JmmY6N2lDaP2DZttHNk/LSOOmtL4Flr3XxscBeRIZA4547r7/8U8LuBvwT8SuD768//6Xnya3sQG/Gg3Jl9cePoCs3H/llO3dIX9sVvm9ikVvuVK1d46623uHPnDru7u9y7d4+9vb0wM3R9fZ3XX389rMx4fHxMmqYh5nw4HHLlypUQ8aLcubdousHBaYFcOX0dxSgdoyFk3W4XEQlhi4PBIDzHZDIJow1bZ2rF6xLC+sy6IqTOjO31elRVxeHhYbCkT09PQ1mV8nn55Zc5OTlhb2+Pk5OTkFYXH9vc3AyhoVZJ2xGWliteM98q7tjnose0npybTxKLZ71C+0baz5JPum83ENgf4Fw7rTkSN1y2MXbErFzSSG/ucw5VE6ifVBguDHn1M2/yM157iw+3H7G9t8Ph0YGPNa+XHhA6cwdlIlBvppEkiY8sSTPS2vAQEdIspZt2IJV6yd+UTrdDJ+v46yUhqbkM5cJnlYS1XpLE+wAq50hSqKgoypxpDnlRkmWC38RJSLIu3a4H67JeUkCKkkrKuSO1FO8EpjYKnaMqS4o6+qQsixoo6/qrlWfpwOXeiXt0nHFwuM/a0pDXX32NnZ3HjMbvk08KYK67w6gpMt3VF6H4P1/jp7lU8Jl2PtNHfL2FUcmn6FC9DvyF+mXKgP/BOfdXReRHgT8nIv8q8D7wvc+TmQJqDOYq8YzDtpmLdhhvHbM2vT0fg7+lekT8ZI67d+/yTd/0TTjnwhIBH3zwAaenp2xsbPD666+ztbUVwgezLGN3dxfn/OJgq6urwVLXe+zs7AQqBjwAKYXR7XYDACo9YkFdlYCeOz09JU3TxkQqnaBk60ctdgVDpXx6vR6Hh4dMJhOWl5dDG1RVxcHBAd1ul/39fW7evBkieiaTSYgSWlhYoNfrBe5eqZxr167x5MmToEzskgQxxWa3ALSO1XimqnV6a7u1WfC2z3zMhcM+0b5txdImz5N2fk37uaY8hfhtcAJCN+tya2uLb/9ZP5O1K2v85Fff5eTglOlk6gE2S3GVB9nSOVJXzbl6mQOWJCmdtOut+2y+lksFPmwxVZ9N3Y7Ox4qLkwC0Vemo6npJsowenjrvdDKfV1Uxm86QpKAoU1JJqFwd+lrTQN2spCoyT8uIg7SmYwJwKgVYkpdemTjwE6+KgqqehRpIhJpTnxXC6HTE4eEh4ytXuXn9Jq++/Bq7uzuMpznz7QLVCj9rUTuUI7dLLYhRCkIrVMejr/qgPBPWvXxscHfO/RTw+Zbju8DP/6j52egKO7vTcuj6qQog3pxBFYLmYYf751l9FgDsfdM0ZWNjg89+9rMsLi7y5S9/mXv37vHhhx9ycnLC4uIiL7/8MhsbGxRFwePHjzk+PmY08h1BLeB4Awqdzt/v91lYWAhhiTbM0a54aWPPi6II0TIK+KPRiKWlpcaCX3EoaZIk9Ho99vb22N/fDyGWqgB1HXkFbHX0KvBbhXDt2jU6nQ7T6RTnXIiXn81mLC0tcXx8zGuvvcbJyQm3b9/m8PAwjGZsu8RtrW1in1XbS9Mr3aTHY+e3VdqWdvuo8kn3bWtex6Nstdr8D7xVHRkt4bWuUTUmMf1yAi46ZvDccrvOT89fu7LKd3zHd/LWm6+zt7/Dwckxs3xGVZbgBJf4pXcTSUgTEJmvypgkSXCcdrKUrJfR63Tpdv2yvWmSUJR+KzyX1CO0eoejUB5J6vA+PFeP+ls6lGmKw4VonOAbq+qRe1L3C59NmLWaZAmZSygF0hoyi6okrcSHTcqc0nXO+xTK0s+q1XDNRBJEVCk4KP2Kkqenpxydjnml3+Nzn3uLew/fZ+/whHw2X+PGtWlhaVb/mX4QRmFzxRIvDWHl+WDdy4WYoaoSO8kUwKFJt9ghvV6nYicsAY0wO0vFxBaeBeAk8SsovvTSS1y/fp3xeMwHH3zAT/3UT4VJS0tLS1y/fj04Fu/duxdoEbXUl5eXA02hSxhnWcb6+noIhQQC7WGtT11+wC7fq+A1mUzCui+69rpSODDfDLvX6zXqTC3rhYWFoAR1BLCyshL4/CRJWFpaCmXe3t5md3eXw8NDFhYWGAwGYRRhRxi6MJqIcPv2bY6Pj1leXtb11Bv13capa1ntJh3WUo+pM+tb0PTafpbmeeESc+Qw59VbHarWejTpn/o41nl6FhYCgCD0uh3eeO1Nvv07fiaDboed7Ufs7+xQ5oXnpCWlNsv9zE/x/LnfcSnzAFhHwaRpRq/r/UXdrEOWJcr8U7gSqrp9Kr/MrtTKpaCeOeqAynPeST2yJfHrrutSv+oglbpuXL1tXlnri8q5uQM3yWp6ygEJncLhEkjcfHap58rrZX+LekazAr5r2sYaAz+ZTDg8OWY6HXPr2gYvv3yX9997n4N8Zjj1Fn/HeW0WKDPXcoygXJpLGpgRQMNJ0y4XBtwtuKolBs3JKtCkVPQ7nF1aQMUCg15vZ7eqWCsQfITIq6++yvr6Ol/84hd5/Pgx4/GY1dXVMMlnMpmwu7vLw4cPA0BrLLxu0jEYDMJGHPosGluuYX1WqdnRgy5HoGVS+kOt6JWVFVZXVwO/rrNz+/1+CL1UC12vv3PnTsh7MBhwfHwc9nBVPl0VhnLs3W6Xhw8fUlVViLvX8qsCzvOc/f19BoMBJycnbG1tcefOHXZ2djg4OOD09DS0ZwzoVizox+fUEtc0+lyWV9e61Pa0fpYXJrZvY95/AwRtr2hMy7QN+efp5pk1FEJI5HNIUuHKtet8x7d/G5tX1rh/70O+8t59tvf3meUFflMKwvIBaZqSdjKytEO36z8lEVKEtNOt52r0GPR6dDoe+HF+rZlUfJQL9SJfRVmSIJSV0FF+pyI4M7NEyBFSB5Wf0EqaQFkJiVR+hUivc6AAwS/9q3Wa4GPwk4R6LZoM+j59UZbKcOPqDVOl8opGxM9sFUmQpEJc4tNWyr+XlPX+C7v7e6ytX+GlWy9z5cpPcHIyoihL3wYyp1cCf/9MpUwYvlm9cLbtI/UdonTOp/guDLifR6O0zVS1L35szVvrzfK7sYPVigKanhsOh7z22mu8/PLLzGYzHj58GKgLO4lKF946PDwMFvna2hpXr15tTAKyykhDH3U1RrWO7T6iuieqBTE9b59HRwSWfhCR4BzVOtK1a5T3t1Ey6gTVLfcU4DWGX591Z2eH0WjEzs5O2AZQwylVeekesA8ePGA4HIalkN9//31OT0/PzC2wI7N4FVDrC7GjLavkVey68UAIE71Q1ruRMzNUaWJAcLBJbW0bZWidb7r0a12btQ6Zj+nFzWFBFUNv0OWtNz/LW2+9wWSS89X33uP+g/cZn55SFXlNCVRUpQfALM3oZh2/XWS3T9bpkNZ0SaebMRws0K8pwU7SQVIoihIhqcvup/i7ygWqRpKkBm6/8JdU3llbOep0jqpyHpSdtqPOPcU7VHFUFeC8FS8iSJaQ1Xl5pyxkeVL3oaIeFeqKkviypfXiZ0VKmZQkLqm34kvq1Sq9tVwUFdPxKU92d7mycZUra6vcuX2Lx9vbnJ6emHbwZWtlaBT8zck5Hs1pmfAe1C2HzPtAyN9Zpd4uFwbcY8eo/R7z8Qo+bVEuFgCAxrK7MSDEYK9c9sbGBp///OcZDod88Ytf5NGjR2Fyj4gEQH3y5ElYXOvmzZvcuXOHpaWlkHeWZZycnADzUEY/ScJ3tMPDwzArVYHdjkwsJ61l19BJDXs8OjoCmvuaapSRBUXl9NUxq8rUb5nWCaGZ6hDVOlbLV5/5+Pi40TZLS0sMh0Occ4zHY/b39xER7t+/z3A4ZHV1lc3NzbAWvW2beJ0f2xesErXH1B9iaR47ArB8fWwwvCiRMz/Os75tIrPyoNjBuXG+Re+1swAj5nSNJlmacHX9Oj/zm7+JxcUl7t17j3fff4/HO3vMxgUVpbdWEcrS887e6d5h0F9gcTig0+3QzTKSNKHT6TLo9kg6Gd0khdQ7ShOpvOUMFHV5K+c3yqCsIEnme6yKIGltkddP5upt8JIqIUk8uKU4XOpXhkxEakvbn0sSD8iZpJD6EM9ut4sDqm6XXiejKivyIqco6qWC/bpnfuOQypGWJR2lfpPasHI+nh4HpSsYjcfs7B/waPsxL924za0bN/nSyrtMJmMqpz4iqS1q1cC2jd38t5u3ka/xWoGZY4hZRRKtI02geX0dWO7BURO9iPoi29mjdlgPc+DTz1gJWMdpHLWhFI0CzXA45NVXX+XGjRt8+ctf5t133w2TlpQ/V1A8OjoiSRJeeeUVPvvZzwYLeDqdUtQ8XVmWYeKNVS52olav1zuzN2oYEhtnr9aPWuw23FH3PtUdlKqqClv0Kc+udIw6S/M8Z3l5mfF4HEYYNlpJgXI8HpMkCWtra+HZFHyPj48ZDofBkXt4eMjGxga7u7vcv38/bPO3vr4e+Hm19C0VZmcSxxEwsYNdy2jj5fXTLiameVws611HLWqdS/1Z22mBhLdzPAnWm3W+hbj2s+N2YvT3dGCPNz77Bm+89ionR8e88+Uv88G9Dzg+PKbIp96q9btTIAl0s4yFhQW/3eKgz9LCkH63W8e+d8hSIev4sFxJHFXly5W4lCxJqLKErPDx6k68LV86f4/KObJaeWRpRpr62bKVc/gYSK8g0tSHTgYfXJaSkFBVJaS+D5UVSKeetIiQdlK6vR6dJPUzV7sD8nJKWVTMypyqrJUIjl5e1rSQY1xTNo6671US/AdVUTGdzDg8OmT78RMW+kM2rlzl1o0bHOwf+iWQHUDVdG7HtEzUVnNlHLc3Dce5ArtrURrnyYUB99gpFnPlKvpSKxDZ6204o7XqYf6Sxxai/dPJQVevXuXo6IgvfelLYV0VXS5XQwp1AbAsy9ja2mI4HIbjdsMLmFufvV4vcNb2+VQBqVUPnFEIGsmi4KZrsOun3kOXBZ5Op43Rgo000WNKwVjrVo+rJZ4kfgG0ra2tUKYvfelL3Lt3j06nw8HBQWMzk+l0yu7uLjdu3OD4+DhsDr62tsbR0VFYfdIqsSzLGtFFcYSLHZVZSseGsyqw21FdW/95IdIgU721qcWaGyaa0FJOZs0WmunCJzzFdlM4ELIkZXVtlc+98QbdTsq7X/kKP/H22+w+2WM6nvqIkaoKwJE4odPrsrS4yMriEp1Bl+FggV7d59LUrwYpSYq3zJVy9uDjFxNLEHH+rwJJ6glQMh8Zalx7mqWkcAkccAAAIABJREFUNV1TVSVp6S36LEk93Dpqh2k9g7lKkbQiqSofe18JZSJkkpJ0vMHUzTKvSIqKWZFQ5iVpkVJV85iTqluS4HBl7VytfIy8KytKEaq6PYqqQqZTvxT29i69bo9bN25we+sG791/4JcrLquapw8NVVvronT82a6h9Jurx2kmnZ6bdyM9MW/fp8mFAXcLPAoW8UtuaQfrONNzVVUFC9aChAKBAqcqgXjXnn6/z1tvvcUrr7zC0dFRWPlRgaOqKvb29hCRsICWBVilVqyCiicnKYDr5CUFND1uFZOuma7Wuqa19M3CwkIIqVTaQ8uk9VRVVSM6R8ukIKgrQapi0mvAO5Z1qQRVSp/73OfC0gvOzSdmaXTPgwcPmM1m3Llzh+3tbXq9XgjXtCMLHUlpvVlwtu0fO0z1GbROLbUTK4KLAO6NtZ+e9j5aq87hF8WKLwjD8bbLY9aXmh0QOt2MzRtb3L15jd39A378J9/mwf3HnJyOyMuijvt2aBxe0kkZDBZYXFlmeWERyRIWuj0ky0jEW72lE5I6wiSpAd7f0yufNIHKiV/1NxOqKvXruyNkSeI3s079Oi9pltJJfZRO4hxFWUEidNKMoiprm9qPKJTCyRKoUh9SWZEgTtd3z+jWa9mkOArx68+kiJ+dWlWEzZWShKrfY1YUfm2ZEHbpyMsSSfy6NVVVUVAwmYw5PD5kZ6fL0vIyV69eY/P6FqfHR5STnMovCuzxXdvA+liitlPwbgR3hP5yVnU3QyGf3rc/+vS9T0kUbOyLrsAbz1hUELQvrgVUBcA47DEOp1OAUbDXaJK7d++yv7/Pw4cPOTo6CoqgqipOTk7CrklKeTx8+DAsEKYWuHLZSj8owCrVoWXQYxacFEh15Unl+jW2HOabUmg9Kb1TVVVwhGp96r6kShXF0UMa1mgB0yoWXb9Gwx2Xlpa4detWaLM8z72lVCsPjal/++232dvbwzkXInuU7w/rbzNXzNY3Yikb23a2LWMnuh2F6e8XzbfXBTHfm6caQ3HFgrmL9ExW83O0KAoXfdbXJL49b994iYX+kA/ef8BXvvoBx0cHFHkeJgS52uL0seMpvV6f4cKAbr/PoNdDMh8JMytyJpMZ0+mU6TSnyAvysvR7pFYVZeUdokXp8NGQAi7xlrtkJKlA4pcyUGu8m3XIag4/y7p0O3363R6dTpdu1oUavJ2DqqZ6ihKqsqKsfBVX1PeCOorGW95+39SkLgcUFVSVuiJSsk6Hfq9Ht9/18fpZpw4B9c5b58BV3rIvZjnj0Sn7RwccHB8wGA545fYWg6Hf2KOtbS1F9jzd0VI1rc379eZQVUtNgSZ2LAKBI7cvfxwtY615OMu5WkpEgV3zunr1KhsbG+zs7PDee+/x+PFjut0ueZ4Hy9mGJ66vr/Ot3/qtwYmq+5QCwRq2oGTpnzg6xFqdqhy0TnQ2qa4gCTAajXj8+DGj0ajhIFWreDKZhGe2VI4CqdIvGi2jlrsqJ1tefQZVSlp2HQlZGso5v7l2mqZsb2+zvb3NtWvXeOmll8L2gpPJJDxPPJFJHb+27jRfW192/kPc1pbm+pgzVD9ZsS/pmffVOE+l+TzNPCD2x51xqOr/xvoHX5drqyu8/OptpnnFVz98j939Pa/0nZlEhlceQkKv12W4OGRhsEC/08HhqZpZWZDPZuT1TkzdrIOkid+cA+9QRZV1WTtGSUnShIwM50qqJKGXZd7xmWV0spRuHVaZpwWpCKWrSBK/w9I0LxDnmFWFn9lK5emTMvd0iQgaS5PUwCcOSlchVUVVebolRB25itI5xKV06u37siyjl2Xk3Q5pXvuUxEfeSFkvNukcZVXWO55NOT4ZI5Jxa+s2G1fe4/T4mLwsaqVRK48GhWLj08/rK4T2s5y7Q2kvn0ieMoJTuTDgDnPr21rW0HxpYy475mDjCJqnDcst3dLpdEJs+ocffsiDBw9CKCMQ9jx1zq+/AvDqq6+ytbUFeI5c9y9VC351dbVR/pgfzrKsMZsV5pSTprdx6jpBSO+nAGiXJVCQV+vditaXUjedTofZbBbKC/Plcq0Vr8+rm3rryKXT6YTtyux2fdPplE6nw8LCAjs7O0wmE65fv87x8TH37t0LETWafxvHbttQ09j2tL9tWj1nn+VCSTTS9vHK87VGYgfbPLmy503strrCvPvhrEPodTNu33mJO9ducnx8xKPdHWaTcT0b1QXqhno2abeTMVxcZHVxmX63i0gGVUlJSZ4XzPKcPPe7Kwme/65SwCUkKX7CUqkWfOV3YcJHuZSSkNbPlyY6S9qvT5N1O37dGnyYo8a/CwmuLChzcFLiCjwnXjmq0jteRbxVXibO0zquQooKR+mdnVXtRHX+utLVNE3N8SSSkHa6dIqCojsjmab1Btu1ylMj0fkNQPK8YHx6yslkzObaBq+/8hKPHz9gPJ5CcHUbw2M+LGv2A+s0bbRz3N41ZaZa/dnYfnFoGevQs7NUrePNDt+tZRcrgdjKsxKDgV47GAzY2tpCRPixH/sxRqNRmOGpAKsORrVOdS11y5Ffv36dpaWlxiqOKqpI1AK2ETA6klAHow0FnM1mYW13Bd5+v8+1a9fCzFJVFtbi15GG5fSV9tF0lue3ClKtcecc0+mUo6Mj9vb2gkP09u3bfOYznwkjFR1VDAaDoGTU9/GlL32Jk5MTXnrpJW7cuBFi4210i7ajtbRjGkYt9rgvxG1t6/IigPtTglhqS64eZjsXIjWUt9bkdgRjs2gwsGopmr8kSVhcWuD1Nz/D4uKAJ7s7HO7tU8z8CBQz4s0SH7nS6/UYLi4yXFwAcZRVTl7MmE5zZsWMaV5TMYXvW0WZM5vlFMWMfJaTFzl5mfvle4uCoqj7fd1eZaCBar+K8zy138fVb/mXpDJf7iDTORkeaitX+olFRUlR+vzL0k+SqoqSsizqczmzPKcocp+urChL3+e8clA+3YFAKlI7bLO54zbBT5ryGI9ftqAiz2ecnp6wu39ARcGd27dZW18ly1If3lnzZoZJ8Uo0ADRn0Fn7gLZz3Q1MX7DK4lnu1AsI7jGFoedia07T2uv1pY6nn8e8qwVUBbW1tTXu3LmDc47333+fbrcbrFrdqk4pmSRJwsxQzW86nTKbzcIeqd1uN6x7boFS01sFpiCnQKv0iPLn8WqWCs468UhB0YZT2qghuz4NEPjzo6OjQPso3aKgq3U0Go04OTkJztvhcBjqeWtri7W1tUDRWH/JdDoN1xwdHfHBBx8wHA65ffs2a2trjUgX23429t3mp+0UR1WpxO1ufTUvmne37/AZrtyS7qYueFq6WM59PG8oXNu4wWde/gyzsmR75wmnpyOKah5pI4mnH9Isoz/osbi4wMrSMt2wcfuMST4jL/ysZ1c6KhSg5yBZlCVFVVJVGrrsQx6VFnGlT4+2kaudo+I5COe8AqhcvXoj87BJpTlU0XmnaFmDdUGe10Bf5syKkrzImeVFWF6gLP36MVWNmGWleVS15S4kiV/CuNvJSDO/omWS+CWN7UjKx+wXTEZjDg/2OTo5Zm1tna2bd+gP6imx6Lip2UZOzArQbZ+27SU6ZhpaaLpy2uTCgDucfWFjqyy20tucbbGjVV94jWiJrT7w/Pjm5iZ37twJKz/qMroKzAoW6he4du1asFDVmp7NZmG/UV3GV8tqY/Ut8CrIW8tZr9NOaR2ielzLbn0ANr2et9sM2u357OJkMXBaWmN1dZWNjY2wZ6qtu6WlJdbW1sI9bPm07nUd+3feeYc8z8N+r9oO85C4pNHWbY50S1Vpu2u5rRKwz6DXvkiJudO2cy07GrT+FteC5K2P5+um1+1w9+5LXNtY5+T0mN29Xcpp7mPma2BNxEeudDsdBv0eS8tLLC0ukiZpTbnN6s0rCu80VVJZ6jXXq4q8LCmKiqIoyQtvDVcK/rW1XFS19VxVdVy4AfKyIi8KyiL3nP6soMgLimKuDHDqL/LgXBT1JtgB4AtmeUGRz8inOUXuAT7PveIpS28Be0O9yYdIkpB1Uu/A7WR0sg5JkpLWlI8oVDv8sxUlk9mUg8MDnuztkXYyXnnlLitLi3OXt8Uzw8qcaS7L2ASrvtnOcau7toORXBhwj5131kKP08Ec0PV7vF5MDFh2ZmjM0y8tLXH79m2GwyHvvvtusBgVMDTCRQFnZWWF9fV1er1eAGNdUsBG+VhfQRzOF0f1aFprpdv1W6xvwDqCbWSMVTbOuUDJAGH2qfLn9t6qEOyeqFYhzWazMCFKQVgdsFaBKXWk9a1O0yzLwuYda2trbG5uhg269VoL8jrCaKNW4hh22572WhXro3lh0uDYm59Ki6hlOLfe27P4/6l7l17Jsuw87Ft7n3Pied/5rFd3s5sgaE9IQ7CnBDywrInsiWGNDMMANZB+gD2SJwI8sQ14YAE0LFAeWIJmFgxNDE0EEDAIDQxSNNB0taq7uqozszLvzfuKG3Eee28P1v52rDg3bmZ1k+xKne6siBtx4jz3WWvtb33rW2P1R5Bit3dfgsXBEr/5wx9h6htcXFzi4uoSfYiAM5BMfn7qpsF8ucDh0QkODw7gnc9GmE4zwxdOFDbJxg+ihiQJqauq2cLvBCaYyjREvWcODsrAiUm55f0QMPQBw9AjDFF1aTKcwayEd4RQeFFyMJChmX4YEEJfHEhCDjZy1Ox9zi/kZKp3HnXlUVcVJk2NplGdnKqmVj2vE6+sJmn7rsPd7R3Or64RQocffvIbeP7RJ5jU9b5gG+9LqJak6Z7bmnJugJG8lO8ftvAfjHF/CEu1RhHYVfyzcIM1asAW7rDbGMM8fH98fIwf/ECbT7BsnvoubIxBI1HXNY6Pj7FYLDCZTHaiTVICSYOkAaZjcM6h67odB2Qdmo1YLS5PQ5vStqmGiGrIXF1dletHB0OHYyN1a8T5zzqWMUecjsrCTSmlneIqi4NbWIRQ0GazKe35vPf46quvSn9Wth60M5cxzLYv0Tq+r7ay2V7HD2nZezT3cHddxsWNfE2jz8o674jefO3w+NFjfP/7P8DQ9/jFi5e4vLpBHDrEkMe/U4y7QDLzAywXC0wmqsvuRIDk8jFkVopzcJLvnTh4DX3h4EqU60S/Ix0xZQOvOjNqhLXvqUb6Q8bnQ07Ydl2Pru8ytDKgjxp5a5+OPN68PmNwuf9DUrx9GIaM74c8e0A+ApcNtRZMOSeQLDQGp/mqulKhvKquDH6elS5Js4zAkGca6/Ua11dv8fb6GmdnJ/jRD3+IxWJW2DvbFLlJrz4wPPfd553XvVj9w2P9gzHuNBa2uMbiq9aYjxOpwG6hi3UM1kBaTju3QUjmk08+wc9+9jNcXl5iPp/vGFdG7jR6FOCigWR0S6jD8tTtK7DFhvmevwVQfsdz4HFz2zT6jHAJedDoMlK2fVAtBGQlgXmt6aDoDDhTYGKXvye+T4dlBctsYpSJVOLwtofr69evAQCffPIJDg8Py/qW52/v6z6Yjoadv7UwjKW4cv1xcPBdLKPgbYfpUAwPYSdsk6o7z3KZ5Yw2bufn5UVybqbBpx9/hCdPHuHt9RW++uYF1rd3CCFp16Jsfpw4LSKqakwmDZpJg8pVhYWirEDVitFoNyc+xaHyTg2kYLu+bKNQNfT6j+elLJeAru+xblu07RrrzQZt26EbevR9h65XRlbbrrHpWvR9hyH2yvABIOJQUWsmX5g4JMSQFNsP2+un40UjfXFepQ68yhXX3ms7QSeZeul0ZgKPlPwOZp6h9C3rJkOxq9sV3r69QhKHT59/gqPjQzjvRqjMfS+8F2oxs9ny2b77/uBWtssHQYWkMWHEOWZMPJRstdPxh7Bs/m5fVSuN1kcffYTpdIovvviiUCJfv35djonVn/wNtwlstW/4NxOxPC4LMx0cHBTDO04M09DScLNIyka0nEXwOJbLpXZvzw0/eJysNLVsnTFsxcXKENgqXhpGygZbfJ77Z2QuIjuJWJu45UwlhICXL1/i7u4Oz58/x+PHj/Hy5ctSVMVtjmdt3KatqrVQi733Y0iOY+u7XlLamxPTP7N0q5RMKx/qeysC6X5LNv5sqzOTDTYE88UU3/vB9zFtpnhzcY43b8/1PgUtsQcSXBbf0qbRHq7yEAEiIZZ8H1zBnrl/B/EKwaQoSF4gWUSRsIE6jnxsoo22o4I0GOIANzjE6CARCNAuTSGrRw59hNSaoA0xIYWAIStFkiZYagOgRUsJfFZcOQ7ntIJVJYwdnAQgKd1S2/5pRWxK2+StJnATnAyI9m6kjIukPIOIKkuwXq9xcfUW6/YOZ4+OcXR4CCcvELa3DVszvXObt/cv39M0us92nJST5bh6z7j7IIz7GKMeG2fSI210T4M41gcH9ssEM+KlEaABOj4+xieffALnHL7++mu0bYtXr17tRH00gIxQ7+7uAGyVEu32LdPFShIwAidsc3t7e8/Q8zdkr1CegFg2K0aJtb948QI//elPMZvNMJ/PcXd3V6R47XmzwYaFZxhh83jopCy1EdiFyywbhyye+XxeqI028UtM/fLyEmdnZxAR3N7eYr1e4/DwEGdnZ2iaBqvVameGZCE2jgN77+192WfYrTEfwzrf1bKjEGMNPYCi25trzkv7tXs+yRp9MqENVCLbfQgUajk80lyScwlvLt/g5uYacRiQMlCys3XJBjJocjMNUdUR85rinLa3A6CMl6zoCIUtkGcaNEplRiKqRqkQDpBir6cbk1IUxSEOAX0cEIIW6A19hyEkIAIDHEJiAjZpwZQ4TaIOmlxlc2xJ2tijdjoLqetanUtIgIPy1tNWKsF5PX4mhYeM2YdBE7/DkHITj/F94PQrIfYDNu0G19c3WK/vcLBYYLaYaWcoEa2OfbDgaFc7aDdytyANxwk/SlsH/w4T/0EYd+A+lGIjceB+SzYAJVkH3G/WMX7YbZTMaNN7j9PTUxwfH+Pq6gqvX7/GcrncgTOo0cLolBADvyP3nZWeTdPg7u6uRPO2CnO1WmG9XuPo6AjOuVIZaiUImqbB4eFhcQq2F+r5+Tmm02k53p/85Cc4ODjAMAy4urrC3d1dcSTUeWmapsgKs+CITT1ms1lxGCw8os487wMdELc7n8+L+uR6vS7QFK8XjTGLpG5ubsr+uq7Dmzdv8L3vfQ8HBwf3KoRt5G9zAXSAlsXDa2IT2NahPzRT+U4Wcwx8WyIzsbgsStKs2OoS4ZmiFkbskO2mt1YCEJW9/ej5J/jeZ5+h7wOuL1U2Q/OiucCm5CdSwYeV0+2QMoziRLI8r6pExqTGPaYE5wS1rzHECERBHzNfHQJViBTAOzR1jSoHYi4m9CHDgzFBklZ0upQle0WQIhDDkOWHI/oYEEPEkBJ89ID3meI45EYZUXH0KsGJRzNpMJtNMZtMtY9TPt7KeQwxADFiyOfsxCENAXFQ/D9k1k8IPWLG7HkTOBuRfPMSMtOnj9h0Pdq+x2ldYzZZwFcOoVcoKyVyc+4nVLbFa1toLdnInWPAOvxc9BbfQ5n5YIy7xdWB3QSbfcBtabllTgBb2GZfxDbGX8nvfv78OY6Pj/GTn/wENzc3ODo6KlgxcWcaMe89FosFTk5OSpTKaJsRMffRtu1O9EvjTrYKZwJ2xiKiDadPT0/Rtu2O8bTaLy9fvoRzKpfAxhnn5+el+YdlnyyXS5ycnJRGGtwOtWCcc1itVsXBrNfrcj50hJwJ8HNKIG82m5LgtTRTOgUqaK5Wq9KX9fLyEgDK/sd0Uo4FOyMajxNryHmeFvOng+BvPwgDnxdG7iUaTFt8GthNnDHgG8Vwe5Nt276bikNPphP84De+j5PjI1x88xqXV28QugExDlqVyX2Ua+20eAgekpR5krxWc6YkcK6C+AoVHEIa4ALgKwfnPeqY0A+Ap+JVUggkiYcXwHmVCIYIhiFBYtCGHEmplUBClQIq51HXFbpsYCMinGg/1r7v1dk4h1h5DEPOc8UA9SEOzteovcd0NsVsMcdiMlOoZ4hAljJ2sUI/9KgjAJcx+iR6PCGiD7nwKmgilzOG7AshUP0azpiYJRn6Dl27QVNVWCynqJxHL9s7V96Z+7R7F3dnXuMxUCy9idy3v9u/fDDGnYvlSO9baLgZTdOgWniDDzyja2AXu2fic7lc4vHjx5jP5zg/PweAQvlj5ed6vcbt7W0xGmxJRyPNhCnlf4kxE+4A1NCz4vXRo0elyTUjdYuRU3Trj//4j3FycnLPaJJ3z+1TtdIWSvE6NE2DL774Ajc3N/jkk0+KbstsNivVt/y7qqpynGSy2GSu9740/z4/Py/FU2ytR0NPjJ45AyZUec/YXISzl7GO0L6Zl8Xf+b0dC2OGDB2FdfofylKMd5llb6fhIti5BrT4ac9/BXJvGzQT3jkcHCzx/Y+eo/EV3l5c4e3lBn0Iu05CJOPjrDNQSd4EhxhVRyXFCJGoTBMIkkvwUQCfk5ri0EtAJUCnYaoybJBVG51SJp0TIABOImKOzLthQNcNEAGqENE4bbARhqEoVfokaAc1uDE3yB6qKsMnWpjkvUPVVKh8hVnTYDqbYpG7RqUBgE+IUZk/UQJq5xCEzJsMCeeoPYYccMVtCVIiaiaEnrKx50wqRgy9dmZL8FgeHKKeNGjXa0QkvX6J8fnYVW+XrSNPxRuk0XccRfIOo87lvVQCEfmHIvKNiPxr89mpiPxfIvL/5deT/LmIyP8kIp+LyJ+IyL/33iPYs5RMsYlorbEfJ84sDu+cK5RARoX2Qbd47aNHj/D8+fMicjWbzTAMA9q2xZMnT4rhtbOFGLfiYYQVbm9v7zFlyCzhMcxmMzx69GhHgZH/eH7Ey1erFc7Pz/Gnf/qnuL29LRCGhUm4/3EbPwCFkvnVV1/hxz/+MX72s5+VJC9nB0ze0gFSOvjo6Ah1XZfcw+XlZdmu9x6PHj3Cxx9/jOl0WmiSlo1koRPuj2JndV3j6uoKMUYcHh7uQDPj+8r7T0M3Lk6ydMh9MzXrFB5avouxXWCXsj/9kBzqXaem30n5L8p72dmkmbOLNpo+ODjF08fP0A8B31y8xt3t24wfGzE2bo/iWZ6QqBYchaBwRsoQfYTCL1pBrxBOzEYoIGX9Fmx57M7BS2VweNWMiTGgGwa0XY9uaLFpW3Rdi3bYoB86Nfp9j7Zvse5bdH2LtlMYcN3eYbVaYbNZo+026LsOgAZe1aRCM2kwrRqI150GxNy2L2BIqjETOdVJSQusosI7KnvMsRR3AY+UNYCADGuhRPMxJPRhg+uba8TQYz6dYzJpsi4Otje8bFB2XnjfxIwBe7/Hq9Kdvy9m+TY8sT8E8NdHn/3XAP5FSuk3AfyL/DcA/McAfjP/+30A/+BbbH9nQNuCH8uBHjMxxg+txWAZCdtkK7fD3zqnEgLHx8fo+x6vXr0qjavJ0b69vd0pMKKjoGEhrFDljjXEmRkZ08CwWQWLg6yzsLx0VZtTBsrjx48xnU7x5Zdf4vLyErPZrET2NMyEjSykEaNW0H755ZelKtQ6KTonOhZ7XiIqbTCZTMp1ODo6wsnJSTH+dFLPnz8v8gN0HEy08lwIEdHROufw9u1bXF9fo2kaHBwc3EuKWraMHQ/7GFJjRpQ18tyW1dffs/wh/orH9vsX2c62sftakpLYjfVMSg/bNXRbAtVkeXR6jMOjY9zdrXD+5g1u1q0a6dKnUx/+YkgcDZYejIQtK40yAPp/lRGIiRFuzG3rIkIacv/RXFAmmVEjkvHpXEkamNgfELqIMPQYhoCuG3JF6aDJ1nZA12bOe9tlOZAWm65V2mTfZ2aLR9PUmE2maOoarnJATOiD/qbtNkqn7PpMp4wIYNV5QBhyEVVMcCnr2rgaIpkWKdi50jZfkaIqVG7ueqxWd+iHgNlijkkzLa0ErbHeWfbd1Hzf748H6xB0zLwPbXyvcU8p/UsAF6OP/yaAf5Tf/yMA/4n5/H9LuvzfAI5F5Pm32Me95Jdl0NiFD/A4kcYIlA+8xW35OxoLGtzDw8MCqVxdXWEYhtIp6dWrV8W4s4CHUaalRXLWQPqirSi1SUbLw+ZC2IbGz1a4Upjr2bNnePv2LV6+fLmD6dNoWsdHqOarr77Cz3/+c1RVhadPn+LJkydlvzanQUybzBcRwdXVFb788stCCT0+Pi5OhA6NkfhiscDjx48xm82Q0lYXnjAMI3kafF6j169fl2rZcaLU4u82H2Hv6diZ8Z6OjT1/8xAs8+sY2/eXbSVqCWr2fYf88N5b56EUGqEZQTNt8PyTj3Ewn+Pu5havL87Rbdoseeu44XIMzmfRsMwBh4hGt9BqVC1mUshG8YkRL1/h7Nw/dduIo6orVOLhvUI6ktEGV5qQKMyTBAgpICbVhwlRtWK6odOIvtPoves32vFoUKw9QX9bVxWmzQTThgwZh5AC2q5H221151VELEfoUa+WVqhqElic8v2nzQyTaY16UpWxNbak6nxzQjXDOm3XYQg9Zs0M89k800rdu27a7k3dTr7uJczFfD8eMw8tvyrm/jSl9CK/fwngaX7/MYCfm/W+yp+9wGgRkd+HRkD8uxhF0vSArTECthH8mGExFpuyhTF227YIiKyUyWSCP//zP8dmsynGZj6fl4bYrPKkrnvXdcXwz+fzwo4RUcbJYrHAo0ePyuAnnXFc2ETog+dEQ0bOOuUNjo6OcHl5iaurK1xfX+PZs2dFmwVAwc7525/85CdYrVY4OTnB2dkZDg4OcHR0hKZpSp9TXksbETOiv7i4wDAMpcjIVrtyHQAluWp163kszE1wFrNarQpUxuQvFTjtP+vkx59ZfJ731J7DGKu377/LsX3vGJLSAyUZnjZpbWJCN6Akz3g+9xYm2UxOTrxgOpvh46fPUTc1Li7f4vz1ObquB5K2vMsHqdG7c7lJtcuFN147J3GTko2yKGUwIResZYPusvGKGVOG6G8AlcYVzkoEgNunj9aRAAAgAElEQVR+p9ExUInDECNcchmLDxh6VaHse5UgGHJh0xAjnFPKZlNp847aVSoX0DQQn3uwBu2o1Pc91psN+nZARERTacenwWmPV/Gi5ycO4gS10y5O0iT0wwwrf4uNy5KQJcjIycwCHevxDCGga3sM3QBXJfg65zFyAdguRL6LoPP+luvH8WBB9+1Q+NbLXzihmlJKIvdlj77F7/4AwB8AgIgkG4HReAO77dQA3HuYLQvGGgeL2/P3NAh1XeP09BSPHz/GZDLBL37xi+JYaHwJmVhnk1LCarXCxcUFptNpcQh0EjRAm82msFxubm4K62W5XBaZXlInKaHLfZBnTsNIqKdpGrx58waff/45vv/97+PZs2dISamJbdvi5z//ecHWT09P8eTJE5yenmI6nZb+rhbyALbVoTzPb775BpeXl4VqSYPMawNsW+nxHjGStzDMer3GyckJnj59ivPzc7x8+bLcE0JEt7e3Zd90oGOjzlmSnYGNcflxZG5hr4fw+F9ijP6Fx7b3fvf3FkSlcUYq9MYEju3ylVl/tCOhUch/iKBygsOjJZ4/f4IkHlc3N7hZrZBCyPvYarSoBozLVZuqseK9QAqNVHfjvRp2B6CPKdM01WB5cUgCNM6jBxUWtTxfEBGiFgm5/L9J5dEHj5SAoVLVRpcckoM24Oiiygp3atBDZq10w4AYQi60qiBOA6RmmqEY5xBDAhDQhwQvQN8HhCxNHJCUogmgqgRBAJ9y1O4cmqpCQK70FkHfd6gnU1TVnSaDeT0yPRO8fQTeU0I/9Fh3LYY2QOsA/BaOec8o4u2lUS/lDxxT+ZLjlxjOv6pxfyUiz1NKL/LU9Jv8+dcAPjXrfZI/e+9ik3LjqItGc4yzArudlez6NtlqHQMhhuPj42JcSS20tEauR42ZxWKB1WqFg4MD/NZv/RYeP36MGCMWi0U5dpsjIHWSUgaEXiiZywgY2DbRoIMgpZFG3TqZ8/Nz/Mmf/Amcc/id3/kd9H2PP/qjP8Lnn38O7z2ePHmCJ0+eYLlcYrlclmIiC3NYR0SHF6NW2rHY6ebmBs45HB4e7ujk8HrznA4PD/H48WO8fv261AesVqtyvThb2Gw2xVmRx287OBHysY1DuL0x7s7xYWc81imMx88vGcH/pY/tnThtbLT1CEtvUY1rdQVJKFE+sNv8Ia+xxdBp4r3D4eEBTk/O4JCwWd2h61rloiuZUKtSIcWAa/WpKwZeBEaUUBBzc+skWskaJMJBhbaqqoJPwCAAAlQ8LB/mMEQkCIKXLR0yCRbiMNSDYtkpIrYRkgKGQRBiRNdu0HZdLibScRsGrVb13qFyFZbzGSbTKSaTBtUk551iAAYgSUIQUQplzJLEEhGSBzBgCMQ6ElLKsEuWIKiaGo1UmC8GLKYT3Na+3J1izPctkpVh2w3W7Rph6JESaZTfIj7IMyR7Lwsbh+OnbCZhu9mHt/2rGvd/BuC/APDf5df/w3z+d0XknwD4DwBcmSnuO5ex4eZDPcbdx1Px90XqdvuMvpumwdOnT/Hs2TM453B1dVWSqc65UoF6dnaGzWaD4+Pj8vuPPvoIy+WyRK+sSiXvnYVA0+m0GGcabAAlUrUaMORk02CSV05oyXuPg4MDAMB8PserV6/w4x//GJ999hnu7u7wxRdfIEaVIzg+PsbJyQkODw8xn8+R0rZClvtlEpfyAOzwdHJyUiAaGuS2bUv0bq+/hUqePHmCr7/+uiRKyZl/+/YtVqtVuR/MEdhWgbwG3J81zNZYW/yd+7f3ffzewjQPYe4PLH/pY1se/CN/VGbtaedRTbL9bIy8c337uWT8+dmTj3B6eoIQE+76FsMQFR6IdAYawTsD5CqsrLx1X1WocrMKpFicQO08gkuoot77SaN5k5gSQu+VvhiSJk0BFQgToEqqrOh9jYn3iJMGHg6r6Ro3Nw0ucV0IEH0YsMnJ0xB6DCkBQSUJND/gcXR6iMfHjzGraySvLfsitKl2jDpDqZwWRHkn2iVKPGrvAHZ6SgESPZzT864dkCqHSdVgUlVImGO+mGFy3aCqHKRnFfDufSuVxTlJMcSEtlfGT4Fuvk1skfb7/Z3Ifbvne8eyb3mvcReRfwzg9wA8EpGvAPw96MD/pyLyXwH4GYD/LK/+zwH8DQCfA7gD8F++9wi2+ymv+yAZa9BtZepYxhfYT6Wz7euICbOAh1EqKynJFLm5ucHFxQXOzs4Kxs5IdoyV29+mpNM7q71C45ZSKgbTioA5pw0/RGSnhR6wdXx0Grba8+LiohwDFSuPjo5KBL5vRkN5X26XeY3T01Os1+vCWV+v1+U6MwFqZxw8/+VyiR/84Af4sz/7s517x+iaUTobflMqYb1el/tkpRp4jzlLsMlwXhdG9Xbs2PdWsuIdY+7XMrb37xw7T7HaCY3SbbSWzINcbHHaOgTG9PrAO8zmM/zgRz/AfLHA5flbrO/uACgsEqF0vgyil6m+iLa8q7xK3zolsSsWnY+ucjlBmhxEApyrMZtNMJnOgJDQ1x6+HRBij64X+L7XIp6YoLRCh9o7ON+g8h6T6QSH4Qjz6VuIc7i8uUCMSXXYhw59t9ECqmz1BA517XFyeoLf/PgHOD5cAknQ9T0CBnR9gEOfaZhaPevhIXWtssIuoa4aKK996zAlX4NYcguA8xUmkwaHy2NczW9ws7qFrHNTbezeN0CTsd5rA/A6a98gJ1uRgAft8Ah2c7yfBcnRe68YvDXv6R4cv295r3FPKf2tB776D/esmwD8nfdt86GFWDcXG33TUI8feC77WBPj6J4GY7FY4Pj4GJPJBC9evMDt7W3BggmBsJjI0hQ/+uijUn7PhGjbtjuVlra69e7urjTPtkZeRIqsADF2ADuOpW3bkuAlVFTXNZbLZTk3slxotA8ODoqYmGXe0Bjaa0ZdGCZ1OaPhbKTvexwfH+Pw8LBASpx9dF1Xzufg4AAxRnz66ac4Pz/Hl19+Wc6dHalotFerVdGhoXMdJ8JtpG7vIa/PmO5ocylj7J1j4qHl1zm2t9sZ2fWUY3MRk6TbJutK5D5yBttc6pZ77ZzqyXzy7Bm8JFxfv8XFpVYuM6lpwZ2dgNJpJakUVowqHyYkeEhJDEruleq9yxK5HsknRFSoYgKGCBd1phBjzPTHiBgHDEGhD3EV6qZG5StAYm7R16LvryEpYegjhkCapZ5j5QSz2QTPHj/B06dPMZnW6NoWWAObLiJBm32ntHV0zgtqgUJNHqh9jZAChn5AKD1V1QgzSdz2Ayqv4gTTSYX5dI7aN3pdJEHocPIVlOwdYkwQJM1f1C5j5luVz/cPjG3iOZnBUXLraWv08+rvJeB893qoeSEuDOxGYnzYARTjyQd5H+Zqf2e3S8cgIlgul6XI6Ouvvy5RICNYGi6W4t/e3gJAwYw5E7A0QhpFRtV932M6nZaInpRKrk+jTCPsnLsnNGaNlXNuhz7Ic+NvWHHL8wKwIxHA2YJl69CZMrEKoETmZ2dnePLkSTHCVhCN0f84Cf7o0aNyrtfX12UmAuzSIn/84x8XeMbWDtiaAjursBAQv+fMYwzH8LqME68fyiL8j02smiVh9J39fvw3sDUeAnjvcHx8ikcnx+j7Aa/fXODi8gp9GJTTnjdASIcbdc6hJhzj2RRa1xEoJo+M75fLKSiGcQjkw+fZhwgESof0OYUbMyauujRU+lJZ4tl0julshqpuIHAQl3aSFN4JqrrGfH6QZ6VTeFdBoHo22tYv5VlCMsli7S41nTSY1A2apoZ3FZrKo65c1nTP1yKpcFjfK2UyxohpPcHBcoH5fA5fq00oMEl2fnQoKSV0mdHTVDN4qcBsyAOj4P5H5b6n91huMYb+4RU/GPmBh7BTm0SzTIjxdNtGcJY3bTFaGsGzszMcHR1BRPDmzZtS4s8EKiGNYRjw5MmTIj9g4QBihDYRag372KjZZKtNUNq+rFyPzBObUCRUQwyfDoWRNY9n3PLORsS8djxfa7AtQ4kOgjOW1WpVzo/sGNsrlY6NSo+2GQh/Y2GW3/7t38bZ2dkOJMR7aGdv1kjbMcFjtQ6c936cWP3glm0uryDmkJJCNdNvk1DdCdnvb5BJWO8cTo6OMZtpMvvy/Bvc3lwj9T0YQ6aUcYLkChaseHiFpqrhfKUSBClXpjqFI5yHKjVq+Q9SElRVD/QeEhO6oUPoMwRD+WDvUeUoWsegoIsDIA5t1wF1dthCnr0a2xQFKWY4QgTIbQAnE9WOGUIAELBuW23JNwQMg0oCpyhwXlOR3nlAtoVJznk0EPQCVFkpkkVXQ0wY4oDUKU+/bhrUXns3HC5nOL/waJHHVeS15D2I6Lsed3drhG5AM2lQ106TnjuVru+OtbdjYHurtxCd/W1OqO4fEGX5IIy7jZr3GfExG8J+ZyGYfcJTdnofo8oAHB4eYrlcIiWVpGWycrValYiSODyjxel0Wlgk5G475woMMm6ZZ4uXGDVTU4aQDjF42/N0fM6MzK3h+t3f/V28fPkSTdPsNKzmPvi3ZcTQkFLDxipfUqKX0BWv6z76JM+VDmEMOVFzhzMUUkatoZ3P58WBWYEvO8uy69vkqP03FgazDtT+7sNatgadz2VBXMbQy72E6oNbAwSo6gqPzs4wqSdYXd3izdtL3K46dJzh5uibipNMqtZVrbPC2RSTutakqpOinS5OlRV9VmaUpHLALnPAYxKkQbnl2noPqKoGk2pAK33hwocYgBTQImHS1hA4DINK7FKgTMEN6qoHhWQAbVidC5SGoVf5grWyamLWwEHG20Vy7sA7VJpUyMfl0GVsu3e67SQBvYhq3qQIhAF9r3TLSZ7pzuZLNJMpnFNyAHXjY9JrGVLSoquuR58GHM4OMZvMdWYAhyBBHXga3/DdZV/ylPruqm2zHRjUt3nX8kEYd7uM4ZRxdEcjN8ZX90E59iHnK5N6jDpXqxUeP36Mt2/flv10XYfr62tMJhO8ffsWbdvi9PS0FPNcXV1hPp/j4OCgGESqM1r4B8AO5EAuu03CWhkAK4FrVS9ptOg8yDaZzWY4PT0t69HBsJjI7puzjZS0VZ9l4nD7d3d3O/i6VaIUkVKJyipe5gs2mw3qui6SwhQQY/ESgJ3zbtu2OMV958nrNsbLx8nS8ZgYO/8P07ib2bfsfjZ+9svrXsueCu6ObKSbpsGjJ88wrWu8vL3FxeUb9JsVQq8NLxRH1w2yc5KvPGazKZazOWbTiUIjRW+YsIVAkhhd9QAEB7iuJBhjCkAUSJ3xbucglYePCUm0IjTEiGGIcBFYty1CBIahx6br0A9UWYVi/SFoVJ1r/7W1n0NMobTga/sOfT9spRBy3Oud1+Smr+GczlIqr4VKVaow+JhbAyqU5J2yaBAHBAnohx6u9+hEoZtZM8GkUa0YtSvb6lw18IIYAvpuAELCbDbBYjEHvC/3R5Dey5rR+23btOTfcUb3Sw7lD8K4jw01sGvE+eDvgzjsdNw6BhoMa7xoQCnb+/r1a6xWq2JwuA6dAlUXyTlfLBbFSLEwqGmanQ5ITKYSxqBBJ97Ocn5CFjSejIhp4MfwAo1+27alWpUURu7r5uYGwDbxaA0cnQk157l9XlfOTkhPJLySUtrRsvnmm28wDAOWy2WhjTLBent7WyJ2OhvCQIRmaPTPzs5KDmG9Xt9Lptt7bFkyFku3UI2N7sf4vOy3jt/NklKJxoDt40tMO+V1zEs28Fm4SrD9riRU1aBOpw1OT08QHXB1d4XLqxtsNkPBurc8eqgBFq2YnsxmODg8xHQ+RZ2bY+gOAmJyqFyeP8SEkLTPqW4lGi+kDJUQAefU2KcUEUWNdIha5JP6CMTc2CYEdH2H9WaNzabHpu8xhB5hSAhJjbwAmXWS4ZMs9bvpW3R9p4VOQImgG+/VaTlfMooul9wqIq/HExHLDEbzBGqsw5DQpx5wAp97rsIrLJQb+hV4S68JkMjASQOSS2gmiu9XuXsVzD17cCSOE6qctaXt6zip/r7lg0mo0qCOH2Jga3BZGGQj4THuah9wYBe3JwWSEeerV69KoQ2NwWQywWazgXOuVJoeHR3hk08+wbNnz3B4eFg47NbxWGyc0MjNzQ3Oz8+LsiIhGqtVQ+PPitWxUbfww83NDV6+fFmqO2l0AZXQtclRXk++0sDSGfZ9vyN/QGzc7tdGwIRyWAMwn89LZM9CqaOjIzx+/HjnvjLK5355/nQqdnYBbGc643tvHa6939aJ8VrYbbyPDvnrWHb2blAZfVOs9PZru47BXezfijZsnxHvHCbzGRaLOcIQcHOzwu36DsPQq1qjtlAqiUznHOpJjflsjqPDIxwdHGBa1QpniGTcfbsf4Wchs2iSFgeFpOqRQFZcTCqk1YeQk6iqJplSzKqUqps+DCrl0Xcd+qwOqfmqATENEMQMG2nUzQT6EEOO2PPMMkMyCSkfN6tuU3E8Ccgwj7JwKBqmejbqQLxz8KIOdghsLt8jpADvvM5onFYOp8y0QUoFoklQhxRDhBePptZI3w6ABw27mH/YGQ7mft/3DA8nbHX5YIz7PpjFMib4kFpDYKfm4whtrG3O9yzHd87hzZs3RdTKrksDRhGsTz/9FB9//HGp+KQRplFvmqaoQjI6tRE5Ded8Pr93rOPEpnVSjH653a+//hovXrzA69evd7RurDSCdSSM4G3hl9WO30crHUf71mCyTSCleulsiZsvFgucnp4WBzdm2fD6kpFjtXa42HPn38A2UreSx/ze5lbG+RZbL/BdLbIztndfd59k8xmw3yik0To5cocIau9RS4Wua/H26grdusvXJSdRAbA9ngpkNVgu5zg9OsbBYolJM83t4RzD3BJR6nXMlcKAslPSVjQspW1EG6N6k2EIiCEghAR9HHKTbEnZOQQMIaIderSbHn3XahvAsE2mknJZ+RreOaQcmPT9gH4ICIlUUYF4MnWQi4o0olYoKeboWrXX+14bhcSQgRBx5dxFkGcb6qC8FzSV5hdS1j8eXX29TlnTJopJXhvX/mCIkR54Hd/vtP1wu+kPPKHKxSbGxsnQ8ff2cxthMrKzjAtitDRMNDzEngHsSOeyVL7ve3z/+9/Hj370o7JdYsuMPsl+ub293Un0NU1T9sdEZ9/3uLm52TGaFkKynxMPZ9SbUsLr169LQ2kmMdkUg3RN25OU0A9lFnitCP9YI24lEnhudn0mdykoRh13cuoJV52cnODo6Ahv375FXdeFPWR126+urgAAy+USi8UC6/V6x2Dz/G1CeAzRjfnu1iHYz8fb/S6WbVu8bcRukBkTpemHnH2nHMYnMcZklEfg9pwTzOYzTOYN+nbA3e0K3dBlo0v4J0NxXrVUlssFjo/OcHp6hoPlEv3QQ1LEEHoV8sqRsCZXNbLtfUAVPIJExbCdA5yHh+LXDo5S6Rlw0m5KvhKEkBVJRTnnIsCApHrqsUeXssoiAoDt8zxpJphMG/jc5i/GHCwkFSarqzy2JPdJdaKsG2wLgwhpsW2eRtzK13dOUHlBVQlScghR182THEwmMyxnC9WwkV0aIlnsCciyyRGVONReWW3a13V0nx9aZPS6b6Bk7IaNQ961fDCRO7BNhI7ZEeN1LL/dDnROxxkl0mDb5CFpftxOXdel6pPQA53FdDrF9773vZJItIVEIYRSwfrq1avCxOH+GR3TQLGVnT1m/m057vZ8aGDrui5Y+NixER/nb6+vr3eMLnF6a9AtpMT3nHEwsn779u296l8WdzEyB7ZOkcd0cHCAH/7wh5hOp+UceF8oAkZHYymhXNfOIKwWu70247FBRz6eAXwwi2XA3IvcdZHReuUzwNDgZPeLsqKqLM7mM8yaKcKgQcRm0yHGIcMHaieU9ugxmU1wcHCM05NTHBwsUFW1YuYpIQVob9LcfDolSv4KKldpchNeuebOZfEw5bYTM05JYQpEl52KUiPhKjhXQbKejRpr/Y3n7AsegIPzFZpKCRCVq7T3aa+CYjFGROT+rt7DFbkEVYvkhGUb3Cosk4LKFKTAjxX6SfAq9FXyO+qUvfOYTiY4OT7F8miJqvaQ5EyuIztMUbgmDAqfVHXuQyvfvs7ivm03DJudhGq6N372LR+UcbdTd/tQ24jMYvJj/fbxFHzf1J1QAA0hozu2q7NMElINx4VTIkr5oyb8yckJDg4OdgwhjRFZJzbatJAMz4EJWV6HcTKQXPvJZFIaW9trQ7jp+voal5eXpdjJ4t2Mtq2mu+1Ralv4LRaLHSdoDT1hIntOFob69NNPS6Nxey+Y26Aw2nw+L+dv4Rd7jey1tNfD5jksm2gcGIwZN9/JIqO31tgjnx+2X4z/sg/1zmZzQtUJ4MVjOp2hbmr0IeB6fYeh7WFL7fU3Dr6qMWumWC7nODyco6kqAFH54n2Hru/Q9dpHdAgBKWPTOgtQjFrc9ngCtk07QswNMGIEYtAoPKWSLGbX0RgT+pD124P2OQ0g9COoKu2n6iuVQuhDQNsO2KxbxeVjyn1TK1TCOhYAyIVFGVMPgTrxChENiRh9LHz1mBIEscA7qrmjGvQQh6ZucLhc4uz4WK9xne2OSIZxHCAOCRGbYQ0gN/JRD7FnQOy3zNbx746EcTCwnVm8a/mgYBkAO9NwADsPqIVYHmJNcLHG3iYuJ5MJTk9P8eWXX+KnP/1pgWeIXdPwAMDJyUnhwxPvJexCQ0I5AMtzt52MLCecETphG0buPF7LtLGft22L29tbVFWFR48eFSGxYRiwXq9LP1PSMa+vr3F2drYT2RKKYqTO68nPqTfDGQtzG5Z1QkcwTmrzfG2/1ydPnpQGKJxd0HBfXV0VnjupkvZ+W7iIi61tGEM048VSK8dQzQexyHa2rchLxokLLEMCXEZ3xZpnbsAmYh1cBdSNGpph0MYRfSEd5F/l/dZVhcm0wXympfUpBWz6hE27waZT456g2Lo4wRAiIBGlQlWgcIOoSYxDBCQgBoFzKnc7BMXDtTNfgsRYuOgpRfSpz5rrA/o+IqQIJxHOO9R1Ax91zHrnEVNC33WaNHUqQVw3k22Fqeg2U4oISZAiMKQA6RPERSC6cs1jiDkZqnY3JlaaYseC0gkR3J4v5nj2+DFefPMa7abF0Cbk5oJgCVoIEZvVGs4JlosD+KrGjnPOtyuNDH4ZC4KdorXtSNjCb8jflEN7x/JBGfexQbd4tKU7jqmQdrFG0f6G642Le2i0bLRM0a/FYlEMl6VdWuNIqqRldFhDaTXJ6TQIgRC6GYtkAVuxMOdcYZgQEiKPndg4k7k0sKvVqtAYbXRrqZfj0v+xdrtNuHLb9rz5Ox4fz5fO9NGjR/g3/+bfFOVJ21KPMsps6Wed93hWsw9bt2PD3t9xEti+/y4XO0JpplPa/WZfDs2amt3v7NpqMVJ0cFD1xYQ80xqUAz5Ovio84yHeAZLQDwEpDtr1aNOi63uI+Iw5C7wM2ZnUKD1GJSdUg0bKsQ+I0GhaE48BcQjaYCMfsrayCwgpQlJEHwI2bY8Qe+W0J4fKV5BaZwMVZ51hwCYEVGHIUX2Nqq4hriqOMXAWGyVz1wOci0hK1yl9YpNSetRwRsXeieFn9QIIVOoshoh+GNAPEdOJx9HBAU6Pj3F7dY2u77fGNSl81Xc9Vps7RAgOlks0db294fkabHMg5n7vQHV2pmUidwPO0fW8D3n87kc+dqNza6xpfMYP6Thq32fg+c9CCgCKsf7hD3+Ip0+fFgM1n8+LMJbVO2dEbrnnjDbt/qxhJtWQyoo0aGS3WBYMDRQTtjRchC/YMJszDzbxsOwhJjLZ+LrrurJfi7FbnZYxzEEnR+fXti2ur69LUtbeI+tcta/ltjkJj4kNsMetAHktnHM4OjragWbG13N8fAB2DP4+PN6yqayj+FCW3RkK9ibR7n+0G1EiG6ES3yXlljezCepJg8rXcOJ1JhClYOCEf3TfDl48kEQhjBAQhoDEas/cObWkH2PS6B3sMMT2c1DDbFrvAQlbn7K1mA5JO1Dl/fVDr5rrg1Iavc9w53yG+XSOqlEpBKVWZrio0y5NoQQS1KEnjLJNmFKVkcwVnbmQfSfln4jWSjmXdXREI/EYBwxD0PZ8IaKZTnF28hjLpZIyNNTmHdKZyXq1BlLE8ckpjo4OdHsobvU9g8O85P9Inm6Vz7jOtxjSH4Rxt5HgGG+2STJgt0DHShZwGT/wjFj5npH0zc1N4ZuT185Imkb34uKiqDcC2ElS0iDxmK2QlYWDaLhZyGQhHjoPYvOMPln9SYPLyJ375vEBKAJejP5pGG2fV7ttOi0mNXmtrHSCTdDymtnzsdfVOibr4OgICe3QoHP207ZtqfK1bKdxzsVi6uN7O4Zu9uHs43W+k2Unobqd/WQ7CIA2Ke2sU7KBexKy2xUkl6YLKnFoqhqTpoKjKiNyAVNBGJLK4jogOTXOpAvGFFVKAKrBpfTBoHTBlBBDFgdIyLz5TIkEEKJWagLISVgVFAshZmkBIAAIKed/woC+67DetOi6DVKIcBBM6gnmkzmaZoIKFVLItqEPGkX3HYZeZyUJuc2faIIzCVSXJkWEkJBiwhC3/VJTpoFmBj3oXUUEidvIVzaEiLYf0LVrrO/WaDcb9EPAcjbFYr7MEKQzBlf58XebDTZdj5OjQzw6PYJ4Z4w1jfJ7xqMZF7Bwkf38WywfDCxjDcc4WrPRDvFrG1HaZZx0s9sfY+qEWmwhDfHmGCOurq7wxRdf4OOPPy77ptEk1XHfYqNkS7EkxAGgUCMJ4VjJAjoN7z3u7u5KP1caZEIuVtKAxl1ECuWSGLiFn6xTsDxzXjfOOpgXYOWsnQWR484cxjgRDGjxEo+Xv7NSxOzWRCEyW0g2zrOM762dOdjrxcUGCR8CLJNM8su+lhQjAVdgi6+mbS5uPI0fbX2L21faF9TXjXY9gkAUqYDkyDqmBJEE5z1qp+JWTJqmlLR5dY5k1VoCQK3yeZ0AACAASURBVESk0UwJKc+M+pBQCxBShIdC23qsymFH5olnWDyLHcYisbvpOrStzmhTTHrcjTJp4iahc4pqx5AwpAgf88wsJUQRjdShBpuMQxEVKCNmLnRkSHAC9Elfi7hltrxegOAELgFDynBlCOgHvQ6cITgRzKYTNJMp/GajzlORH8QYsF5v0K57HD06xtHxGXxdw7UtkARBUuHfP3ArdSZkPkjm3Xsj/9Hy3Y/8vIwfThtl2sXqsAD7p/BjA2Cn6sSarUSvZVt474tE7snJCZ4/f14iUxrj9XqNt2/flmYTrLS0BqnLXdc5S6AmjM0jjGcW1vjyXPnPNgxn8tfOJFhxyvwA+fQ2suZ1oUOwht9Wi1IjZtyUhNCKFSizjbPbtsXd3V2hh1qoZrPZlPNq27bw4K0jBHYTp9bY23vJc+HrmBa7b8b3nS57Zg47qHmZdlMegFN5/e/+Kfh4mxFdt0bfDUhBo9UkKTfoYHl8dobiVY7X+8wW0UhXlxxkgSz1iCHo74cQEfqAfgjoO2W6DEMsxg25G1GIqUgO6ExBDXqCfhajqo22WVOmH3p1AllNsnKaC5AUMxUwQfJsQKcgEYgx81sy7JLplDFr0/AYEnR1gShME4n987rwam4jd85a+jCg63u0Q4tNt0Hbd4gJaCZTTCa1dnsy9yrGhL7vsO7uUFcexwczeF8p+0YyNLOnc4eM3oiM7/3YrKcCtb0rkv9gjLudZtPg2AKUMcxifzd+uO16Fj4AtkahqqqCs9ukKZOVBwcH+I3f+I2diksWJT169AjL5RKr1Qqr1aokZ/lKw2cj1nEylQaIXG8aaQvLsKDKGnjOWrg+MW1i+TSOFEWjMRxDJxYG4TXkemyTZ6EY2zbQwjXW2HKb8/kcT58+xdHR0c6shPeHPHri7uTEPwS1cftjFo0tjGINg92f3eeHsqTynz2fj9KliQm2vacgBntXTZR2MyCEAZnBV7o6iZjkW1KtlUkzRVNVOoPImDlEIC5lrjerPSVDMZleGHX89kGrRPUZVcsaZbdhisrAC+os56udnbIkQYwIcUDo1RALAFdr2b5zLuPkko11QqBSZNxCLiknRMk310VNftY7K5BIYchAMX89H+LzCSqSoJi8k+xDQkQcIoY+FlnhmIImdf0EvqrKdUK+jkM3oG17eO/QzObwde5qxRSpnY5Z9M28SeaDlDjHs+bdzP/eEcx/MMZ9zIAAtobYJsgYYdoHnQ+0XSyePY48KQL27NmzEsWysQQhg8ePH+Pw8DAXg2xKJDufz+GcatSwmMnSBRmx0/ixQYbF4blY45PKwN3NPXC/dHbcpo3GY4ylyIl6MSGEQt3kfsg7J5wC7CaleW1ns1mBT+z3ZPiwQpeRueW4k865WCzw2WeflZ6yPD86LdJA+b09X14Pvlq5CRvFW8YOsEtHtWPqu19k9x1zdnYNA8zK6DPZ91ybqDW3/dFoVyhvWyGJFhAhSTEYzgvqqkE9aeDqaouXQI2Bc1qJyqRpiTrLenlWRQiOx+PUgDnRwqLKKfNlWlWY1DVm9QSTyuux5dZxKQiSi/ACOO8wrWvUtYf3iu+ENCBBOfYMy0OMGMKAblCRsRR4znrciqazsnYLpXjRQi/1NyZzSQfGdZyD9xWq3MxDOw6mok0TU4KvK0ymGrA58eae5BxT7JEgmGbJhDITs/fdJEfHr2K2tz1UO44NIP+O4f3BYO7ArozAviQpsMVdLd7Kz8eNHuzn3AYf/hi1ohTYVlk657Ber1U69dGjYqgYIVpDZxt8sNUeDRgNIYAyE+DveHw2KcxInpCHTdSS9UKc3cI6dHKTyaQYd7vt29vbIhdAh8TrSafIv3lsPFd2cRonjvm9rQrlMbGJB50qo33quZNyySphG3nb+zzG2fnK9/uYU+Pf8H59GEyZLWqeAIyluNModNv5kyvLzipmyZFdEoRejaCvHLRPBcerSTZDDV5TVRrZxaQ4eswGLGaZgmwkE/eRIY8YE4YYkAIgEhCDctODSpZrshYCeAeXdOxMphNUvkLfDmqc+4jQ9xhij9irwa4BQNSwDv2gs4GgImMx6qxhgED6HptNi7vVRmcezsNXWQ9HZBud5+sUEwo2H/N1DHELW6U82wgUwqRz9BWqKgAhP/8CjfhjgCQtGlNoyyEml7tAqXBa17eat2gMyiAjdO6h9zvDZTtQUsIeSOffksgd2JX5tawT4P7Duw9uoUF9CJJJSXnpZMdYfJcGgSwO6qLY74gd27J4azRTUlVJK5FrK0Atm4bJSkav/Mzy3SmhS6kDOpq7u7tiOImd03iSPTMMQxFAowG1EbO9jsDW6RH3ZwUuIR5gSxGlJj73ac/VRs2UKuA+bfvAcRcrXhcLuVgIbsyasfmDfSyrDymhakNvE6C/Y718fziD37N+uvcuy/GmqGX79YSZxDKldxC4yqNqGvha+4IO2WBR3VCEEgBpixMzOZmVCTXCJ6xH0CAXA0XFyV0STCqF8Zo8RuMQ0HUt1ps1Vu0Gm02Lttc8URhytWpKuZerL6FughpkpU8O2LQb3K3vsLpbY9N2CIEKktprFdCy/a1UMHXXldOOtK1MLQFZorIkZzdVGeeVrzKklMNtSZCqQlXX6lgcoZec/A0JcA5NPYPUKkF8jwi5L3LnnTR/78/WmG392xK5A1uDsy9JZr8ff2b/fghrpSGiEbEROSEDskiIL1s+uYVVLBww3pddd+xcxtujobM4sYhWpb5586bwyIlPWxEv7pfHCahMAWWNLy8vcXNzg+PjYzjndrB6XmPr4Cyzh/vhDIk5Ce6Hzo6cf5tv4DX23pcm2XamQgM8m81KodgYT7fXlduzdQQ8Bnv8vKb2vowT8t/JomHXzkc7KVNh/M3vbIXqfl9wL8GGhHazRr8ZMJlNUVc1xIuJ/HO5vAe8q+CdwgzIeLvqseTIPSndEZyFJoGDUhmVvy5AyrTWBEhI6FPuhJTplRq6asOMBJW2uGtb3N6tcXu3wuruDpv1Gn2Xk6neoe87CCLq2qNpKrQbj0Ekc+OzumNMaCG4y3kBX3tMJw2aLFUQhwiXZRDQ0xkpiyhFhxBi7vkKOEQMSVQyIWjyWZLAi4OvHFJqIB4QV8E7KZi/E6B2KcNAHpJysCcq8pbioLTUxsHnecP2ro2B9vszObu+Ufs335BFk/YPjry8N6wRkX8oIt+IyL82n/23IvK1iPw/+d/fMN/9NyLyuYj8WET+o/dtf8/+yqv9B2whGS4WpnjIONBw83sW3PR9jydPnhSIgTi0pQ/aatZxtG4NBz+3UIClBTKiJV5uE5FjTJoQzMXFBW5vb3F1dYX1eo35fF62a6N2KynM7XF20rYtLi4udlgu9rpZ0a7xwgSwnUWN2UVMFFsnZgulvPd4+vTpDq5ucwCEgubz+U6xk43EudgqYR6/zQnYxbKj3jPWfi1jewcaMoD6+2bpu4lVGLuQMeuSJIRGjF1AwABfVVjM5qq8CMnt8nLSEk61UaqsfeL4rPE17rZ0g8IBCYB3SgmsfYW6zjklpwYNMeujx2xQnUPjq2xUE/phQLvZ4Ha9wWqzwXpzh02rkXs/ZO56SnBVhelkonTDpoG4Kndg0u0PoUfft7jbbLBa3+HuboOhH7TxSFVjUjHKJgtIk8AhZE2ZTNBnkdPWZuREbMbpva/Q1CqFMK1r1FWNpq7R1F4jeVfnQi5o5jhfLweH2k9RTyeYzQ5QNxMFuHhB79v2vahMudO0faPEi9CBvmP5NnPWPwTw1/d8/j+mlH4n//vnACAi/w6A/xzAv5t/8z+LyHs5afZh599j1oT9DtjlVFtjxb8tPk5DQP63pRzSwJCqyL+JE9N5WOXI8b75NyES4s7EwGmU6SSIXVsjbc91tVrh5cuXuLm5wd3dXYGT+BtrSBkpz+fzsj3SJYdhwOXlZWn6bSNq64CsM+O1ANSg3tzc7HDx+c86PZ47r5+NotlOj+uzoIr3ysI/422MF3vN7dix+Qqb17CBwQPLH+KveGy/a9ln8+172fkgbV938mnElxmJq2Nezheomloj6URlRtVxb6oGk6qC8xVqqbTLUGHIaL+hlCzHeptsrH0OKlyFOv/OM+MadW0HQeUruMrDZTHgMAR0/QZdu8ZmvUbbdhh67bqUgjoe7ytM6waT6QwH80PMFzNMJ6o8iYTMXkmqx951aNsOm80a3TBos5K6QVU3qDkLTluoSM05jXkeQyllk75lkavRzCqYUmlSNT8z3jlUXuWHp02GRJ1XB5oLpMQLmkmNpm6wnM8wnTQlF2AhOdl3w8f3X7ak2L2j+IHAjMt7jXtK6V8CuHjfenn5mwD+SUqpTSl9AeBzAP/+t9hHeaVh4DR/zISxuOqYE21Fx/YlQYdhwO3tLTabTcGnaYi50CiTo84iGwAFdqBhtDRAskBYGMR2dowugW01qWWzWNYJt3N+fo5vvvmmKFXSEZ2fn+P6+rpcBxpZvrIhiHMqBHZ1dYWbm5vCsx9fD2BbYMQqUt6LGCPevn2Lu7u70iuWTmXMjplOpzvyCfaYrRPldzb6pmHnOfEe2nyJfW9nHHZmxmvK7x9yEKNx91c+tu/vs+x7NO63X24/B0iZS3a9+1tFjEDX5RmmF1Qzjao1akTWIdcIs2o8qkmlOuil89Jucq5AwpLhB3Y4qnLZvt/OqpN6gwIWKEEnc8yhCdN+6NBuAtquR+gVYw8xyxxkSKOqazTTBtNJg+m0wWwyQ1VtGScJ2r5P6ZghV6v26MKgPVyd4uXiFAqJCdvoPG6TxppQJQ6/PW4RAZIa0lhyCHl2lHMXzgkqV6NualSVsoLE6/XgvUgA6qbCbDHHZDYtjKO0XaU4z+yXy+f3x0je5ujGK9L3F4/cH1r+roj8SZ7anuTPPgbwc7POV/mze4uI/L6I/CsR+Vf5752o0kaAjA7tAz6OOi2WbR2Ejeicc0XW1sIH7ANKw+K9x+npKc7OzgpkQh75er0ukbd1LrZyk+djDZB9Tzomz4n/yCL55ptvcHV1hRi1V+t8Pi/VnIR2SEts2/Ze8pHn23Udrq6ucHl5WaAnXkN73XiN7EygbVvc3Nzg0aNHmM/nOzMPDioeB2c65NvTQa7X69InlTx07pP8fQqicdZgjfPYaPOYx9TIfc7g2xj3X8fYthGwfrf7mv96OHSX3Y/uJd/Mf4a+x6bbADFhVk/h6irvR0EZJ4Laq4TtxNdoqjxuvIN3mcbos94KtVfE5+8dvBfS4YujEIHi1Llxh5P8e6BUgpaxj0FnGkJIXmcKlfdo6gqTpkHta1WtnNSYTWtMJjP4ulYGEPNXJkcRh4A0DAhBY2/v1Fnpv1SKoLZXi+VZW0ckPGZRCEtzELp+AgovHpmPX1V6DeuqgatU04bGPwU9P4hDVU3QNFMVrbHRt+ze4vFUTcyKhSGzc99zPiaNwPrR8qsa938A4IcAfgfACwD//S+7gZTSH6SU/lpK6a9ZTN1CHtbw2IfdRoH7Hv5xYQ2/Y5R8eXmJq6urIlpFg8z1nj59ikePHu1EtQCKk+EsgRALnQL3udlsCu5t28LRePLcLDRCPvt6vcbV1VXh1FN/ZbFYYDab4eDgAIeHhzs0ydPT08LwIYS0WCxQ1zXatsXLly+L4+S5WPjFzow401itVnjy5AkODw9LUZOFdQCUDkq8DsBWf8cqXnIGRsNsDbCIlEbf1knba2RhHPvb8ayA93o8fn7J5S99bL+7zR7uJ8V2bdH+73b+1grSdrPB3c1aI2KnxUOc2kveZxKB97nZivPGKWrD7JRcjvb11TlkJopDynz5IUT0vbbPG3Jla4RAqYzKj/fOa1FQCWq01Z4khW28r+BdhcpXiuPXtUIqVZWbXCjePZ82mE4muc2eMlOooe6gs4YoEZFFTgBEXJYDYDVqRtlzxBzT9rpt5XelXKvSCzYklSDIomp9DFphC2hf19oByZm8ZsoQVAfEiMoDVUWuehrdOnPT33WP035gZry1fcuvxJZJKb3iexH5XwD8n/nPrwF8alb9JH/2vu2V9+OH1D6o1mjve6Ctk+DrOGmY0laXnVH5MAyYTCal6vT4+LgwZ6xBo6GwGLut6LRwBJ0QZw9WnoBOwxoy9mb9/PPP4b3H2dlZUVZkUdF8Pi9QBh0Vo2Jqz19fX+Pu7q70NL27u8PNzQ1ev36Nzz77rBhGJjBpbMmj5+eHh4c7zCBg6xCsM2DET5iJrCPOMvhKYx9jLDg72TI2MTye6Yw/4/HYpDSPzWLsduz8Mstf9tjOG7r3VmRr4MV+mdIOqeJbHLFGjEnZJO3QATFhIlWuoHRIuWGGRssek0mDqq5Qi0NwAp88gs/J0sohieLLWsSkDapVMhfoTaVqhNcuTAAq5xBSQuUrNL5CcgIPwYAI7zzqyqOqvDbXqGpEZHgOQF1XqJspJlWtcr35WH3lMZlOMJ1MMZ1NEZPCO0iaO6hqr7CNVAXqkIyvV+IwOIHPxlFpkg4JDpWLCDnKryqHmAQuJQTRRtkQgUQFZ1Qbhw09VMoYgJ6HU8eUnMJQei8dajfJ0FCF2jXa4WlEhiQHJmGPb5ftpK3QYZNdz+A771h+JeMuIs9TSi/yn/8pALIN/hmA/11E/gcAHwH4TQB//Etstxhp+2Dz4bbr2IjMPvT2N+Pt8ZVdhgAUY0a8+ujoqPC6x0qPNID8buxEGAmzOQWAYtTscdLIWRXG2WxWIv4nT56U41kulzsl9lVVleOIMe5UwzrncHh4WJK0NLKr1QqvXr3CZ599hqZpigbNQ5Evv7N0SBptrs/f8By4Pn9vBccICTHHwPO10JmNzPkZ/1kZCgvZjRc6nn2zuF9iDP6lj+2dBzj/MT58RtblPXZt/J708gh3zfmHoJ2SFBfXUvpAHER0FtxUNSrxKigWBcklILfSg2gFKSWBJeP2TrweQ0wYhoBhiJAqac9SEQxJaYQEPgRO9d2lAnxCXU8wmzaYNTVizOMrDhDRqH02naKqG6QQ0Uct+0cSNNUEi/lC9dNjhO/0OKrKY9JMUVcNfKUaOUMIiFnrRgNz2UobQGctThyieJVmyLAQK2bVqWwjcZb+x5TgckOPqNMRCIC61loVB+2VmkSlG1hU5ZDgiOmPCpDGqFzZp/1OtuPiYW7Vw8t7jbuI/GMAvwfgkYh8BeDvAfg9EfmdvJefAvjbAJBS+jMR+acA/l8AA4C/k+jqvsViMWlgtzTewh58uMewgjUCdl0aCGqhTyYTzOdzvHjxovyeke/d3R1ev36N29vbUmhkm2/w+KwMAimUlPSl8bYFRtZ4WqNqm1ys12ssl8udJt6scLXSAjyvtm1LpyieO3VaWAl6c3NTInh2baLgmTXcNM5WooDROJPK1inx2C3/nQbb9nVlIRalfdnIm+qbPA6Ls1vMnHUH1kHbsWDv9S+7/LrGtj2yBOw0VkoEzUUfbZtge+cjnGTHYKSU4YQYESQBFb/eapYrnq69TIcU4YIWPiFQXSVBwCIm/Q8NU0wBEkVb1+U2ejIIgoQy4xARSBDESvFppVqmIgPQ1BM0swmGlOB9QB1V5rducj2E95lzriSHIUZIZsFMZ1OEIaKqtCOTd4rTN3UNB4cwBGxSAvqIPupMZYgqXeCg6pcOkg08NDrPLKAAaBGT5Og/Y+iSCegiUJ59ShAmZTPDxzkPV3nIwOY4QOWVpRQlaSPx8SgwULnY+20cv8guosHPtr/K/J53DPv3GveU0t/a8/H/+o71/z6Av/++7drFJkktH3rfeuNkKICdSM8WuNjkKj9LKRUWCpkl3mvnoE8//RSvXr3CxcUFXrx4gWfPnpVEIfutjuEDW53K8yBzhhGwVXS0RooLHUVd1zg7O9uJQK3xFZEdnJ+wB9+v1+sCcRDmoSgaFRvJfLHOkdfQGmpG7fY4LX+c1FELezFqt1TS8/PzolfD3/J3TArbQTwuDrN8+3EUbo29za/Ybb3L6P86xvZ4KQk0RmvC9BjU4FhYxsy+SxRfvtuZp+ftJETJ/VSbGeosn4ukDJLKV2iaCl60WjQhoesHOKc7U8rg9kDV0CFDEln0MaEADFJYHKkYrJiph85vYR04DRamkxrL6RIOHkMK0NSAzlznixl87RHTkJ1HzqM5QTOpcJAWcOLQD70eQ9aOqeoaTpQB45CAEBEx5CYluU+q02NMOeIW5+CyIVaOfirH7sTBml7i5TEBMWh3qYCoryHBeyibKClDx3uPetoAkgumhrDLiMnXdqdwyd7vHXutCdWtYeeK+XmzXmLP8kFUqPKB3Jc0G2PvY442/6aBHydUbeTMB30ymexgzN57zOdzrNfrUrJ/fHxcDBIhjlevXuH29rZg3fP5HI8fPy4VmJb+Z6mHVsuckfp6vd4p/uE2CQuReQJslRrpKFarFd68eYPT09NyTcY9UAEUaAdQYzydTiEiRRLAcu4tDm+P2c406HwtJNV1XUkak9ZI57fZbPDVV18BABaLRZkx8B4wwcxon/fDirFZZ8jtjiG58T/rfH+FhOpf6VJsdrHU9stRyG4e3HTvjcHn+dMEeHhUTYVmXmOa+4ymmOC0izYgDjFpd6NYuowJRBhM5LL9kLT/KJRCKM4BcKUxiDiXm2QTAokZ+3dZvlePUQtjHXxdoa4bzCZ9+V1MEc5VqGuP2XyKqvJKUQxBq1GRVMOlrjHPHZRiIQKo8fa1R+3JClJjLRm0TjsXm4VfTH1qdB5Dyk29c57HayFSAu2HK3BUyM/FELThdh97hEHhG8lGuW4qTCdNlkEICCnotZHdFOiOyuO7EJdkLf94+QtG7r+OxSbB+J4PJo32GNJg5EdDbh9im6i0Bp5GiDAAS99pXK6vr9G2bUn22aYYs9kMT548wWw2w8XFRWGwEB5hAtY6F4v9WwNpj5eRL3/P/XE7vB5Wi6ZpGhwfHxcMntWoNOrWudmWgBZKIked0gZMetpr7JwrMwMreGYZTFVVFRjr8ePHAFCi/ouLi9I0ZLlcous6nJ+f7zhZWzfA31rYigv3yzGxbwxZuIbn8aEZ951j5GfbL7cG+z2wajKGneu6yqPJaoUH0yUmU6URDjFCvLJnvNfeqMMQdlg8TjQydw4IeWaw9TFqHFVd0QGVKiGKCCqvxlyL+aGl+zli91nnhcnMpqoxmTaQSo0moLz0qlLeeFN5dEPQgk8HSBY5cwlGpCtmgzwgBtWhaZoGja91f1KhT4Pq6AjQDTw/lyUEeO21D6AWgUUgZclfQPnyScqrQNCnAAGlOXr0QVvwJQmAsLpV0DQTHB0cwHmPFJ12MIFTXB525plAJP0h+yz2Pzv3O0f07x4iH4Zxt4YY2E048vsxzkojNY7qLKuC69IREMMmnn14eIjlcokYI37xi1/g6dOnOD4+LvK/NAzsZVpVFU5OTnBycrKTsON+aRytYWbkz89tnsCeL4+R50WnRodEid6UUnEsVtiMEbqFLsYa8bbJB68vZx1jBorNMdDp8PvValX0a0SUOrpcLtE0TeG30xgfHx+XxiGLxWLHOdl1LUNm7CR5H/5/6t7tR7Ijv/P7xLlkVlZlVXVd+n5jN9nT7OaQnBkOORRnhhLJkbQjeHcESNCbvTAMLATsgw34xfBf4CcD8ouBBfywBrwPAmysDFkC1rsrrKTVbTQz4nC6yW6y713ddb/k9Vwj/BDxi4w8XWxS8GqmNohmZp08efKcE3F+8Yvv7/v7/kKG0WHB8nDC/Htkqf582wRQncJMbeDsOWC7mf7DwjD4LytlDXTqaI7tTpvuTNsKYSlFrBJU4uoGOAjH2guBwBzQYiyX3cUNnRGMiZWUqYupVG2/FYFSMbHS1BgsuuM43wbKWhNHkiofESUxcZzSsgEAq7oSQ6JS0liMoHYl74DI1mxVyqq2R3FCbCOj1HWBrhUqhrRlay3EDoJqY2wlqChCqwxd2ftjKZCG2sUnTKRtjEC7YiVqUuHJUidjlNIYIiJJyjLWg6/LmrqqMJVBaZuhqhTMzHSYmZ2xDCNTgy5xd/vQTj0091Swd/B8e79d/v4iy84RMe4w4SQ3YZhm0CyEWOThD7MeYQLvhNsF5xXPbzgcsry87JOExDjNzMywt7fHYDDgxIkT/lwkuCjQRohbhxWH5PhiLMWYhsFLyZ5tBldDGAcmlZ/CSUTwdOHri0cfGsLQWIaQSzMgHDa5N0JnbE5SYlhlYhSmjtbaY/sywcn3Qt7+wcGBv2/dbtfvL/GJsL/k3OR3wuImYZsOOE3eh/GO5mdHqUnSuzFWQyRcgJvPe/Dlu4e4fCpSRInVRem0Z0habZTcS4WjNErSjbGl8Vx0V8rwRQpqZYtZCyJgb6MrU6c0CuOgFeu510bRUtbztZmwhqquSJRl0aBirCFzVEsdk0QxxiU7xdEkycooi2GbRGFMTBxZtk9kDDq2K4ckjigrRRRrTBTRSWeYcQH/SFmBL1PX1EaTVjF1YiGjyBgqYz11Y9wMFtlJLlJmyq+28QBNbazHb3RAZdRQGRcHQtsM3cig4oh2GtGJW4Ciqg2lg9+m+jKYlaf6OTDq4QiR/p4KqPpEq89vR8a4C/YtODNMDH6TCndYQFI8t9AwhWnq8l3BeUM9k8XFRd5++22SJOGv/uqviKLIZ2cKZCJsFin0HMIswjCRSUcCnc2gZWhoZQIIk3tCJowYtZBfD/jjSgBVOPmhVIEY8dAgyz0Jg7tlWU4lVYVUz3CykPM1xnhOf1ikOwyoAl5PRrJQ19fXqaqKM2fO8OjRI089BVtyTxg0cowQjgpjL+HvhRN7k/4aTgRHwbBP2eFD7bX1n52pd89u03IzZRSUMlPSsJPnwV57pzPP/OwscZI4rzFCRVaDPIotPx1liHTkDbVSjs5H5CAJrAFWzvHCGkTnALtsTluYW6mI2DgdeRWhjKY2GqUjC3tgM2GjOCHFarAYY+ERj0QpY/HuOLL8zjGU0AAAIABJREFU9MiW3tO1tvWWDKgkIo4SWy/W2HE8005pt2eII7sC0UZTRZCUVt9Gl85zZyIpoK28JbWxHrvUXcXgioNoS3vE1XpSdvJUUYSKIak1OVJExOL8ERFJ0oJEORJDRlnkOFWbqb6cyPSr6YAqEyNu/KSrsFLMExzOmC8e20cGkAw1WEIDF3qaTc8thGOAZ74ftpAFIsZRjLYUwlZK+TT/7e1tsiyb8kJl/9CYHeYFN9k/MPFCJYAohig0wICXOZDzE3aOpO83WSOSEBRqwct5CbUxjDOEnn1o6MPgowRdpYU67HKtcmy5/8KQkclK8H45/6qq2NjY8LVUZTKR7FS5Xplw5bjy29KHYcawnHs4ATZZM0fBuIcjsYm8WMdOHv0vk3cYHMd7hcHfBlQcMTs3w+rycebmZq3xxZbXi5OYJE6dqJfFlWNlRcMi0R2PIFGRnQSiyWdWI8UxuIwzysppxMcJSRqTJqnNUnU4iDE1CqfJEsckcWRXFyoIyAozxjgpBJfpKg6fnYwMRCLHG5O2ElqtlHY7ZWamY2uaOhmLOLYZsnESuYknyD412ACn6wg3aty1OQhJi9a79nGEKLaB2zSJaMUpUZQAymWzuiA0xiZqodB1xbgYU1aVL+The9c0+n/qj8kYsUZ8so57tj0fbjwynrtAEfJwisE5zCMP22GMmpAZA9OUODFKo9GITqfjqYybm5ucPHnS49kiHSBSuyF7JMTXxeC3223PkAl1UsRYyXnIykSMZxNCabfbni8flp9rBhnFAIc4vhhWSWwSr7u50pB9xXiG9+iw4HV4r0NNndB7Dr3xMDGpqiqyLPP1ZpVSLC4uMjs7y8HBAXt7e1OlDEPILZzkwxiF8OvD85b7FcZijlowFSZemZqKkU02hkHOZsC0eZyJyXDjPYrQpiJSitmZWZZXV1hYWGRvd4/Y4d1xYkvZJVEMKZhKeckAsEFVVVuaoTV92hk9x2BzgUotAdbYUiwT54knceQvUmMwShQVE+utag2V1a6h0n4FEkXKGkZj4Zpaay92pgrLMY9QpE7VUsnvJymtdkoUJ2Ccg2gkgctNDNqyVSIVEZkaLTEP543HTlZB7r0l2xi5s0RxREqM0gZTW2ekMoYoHkOMBfCxAe1227J+yrIiG+dUukac86nVW9inn+O541YMBoUyoeeuUF/CETgyxv2wBzH0MptLdWDKmIfebDMVXeCZkGctHu/y8jJRFDEcDr1xWVhY4NGjR4zHY58eL8ZCziOsINT05kPvtuk5ynck0NrMdJUJQ85X9pNrEpEteS8euEgShNWowsko9HblXoUB66ZoF0zHO2QikOMLLCN4uExsIZRkjPHUUpFLMMb4oLYYfqGXClQUiq9JC+MOoXhZOA4kPtGMyRylJtipCbw1F4f8HLf+kPfN5gpg12VFNhhT1TVJGrM426GTpFaSV1ujnASiXpV2JeiccVFKgYqtiFhsDbkGizerCbvHBiSFaSKQhwiB2UkmUooKmyRkN0VWOdFN+tpY3F5rUNpmjloPXVFVBqViu7+BONGOx668Z46ymjFKWVpnpGt0XVJUmshAjXYrIuPiE8p57SrQlrF/G+x5GGOLfCtj9XOUcSqTDpaK0pi0TjC0XF6KLSUok2AEpK0WBkVVFoyzAboowEgwdrov/Qhu9m2I44VjQk0O4MfMc9qRcW2aBiVcih/GlgnZEeKZNvHWkEcuhlS+Kziv0BkFImi328zPz3stczFWocESiCaEU8JiEjIZyXVIQo+ca7jaOAxHlmOEBUSaGLIYYjmvEAKRQK7cl/B+wMSzDg2gLIH9ElZNYhZyLnLccMIQqERWJc17srOz4883jKeE8YbmiqBpuMNJNYTAwj4Px0HInjlSTBncMxs8zMZtmIZpJq/PenOTI/lnwe2kdc3BoE9ZlEQqIW53UK3ECmahUU4YDKOotCHShspYB0I77DuJcMJfEUQW546U8UWqtWPFRLELEyqoXF6+hWuM3c8oEuUMo8PyjavoZBF4kdy1gd4oipw+fOwom0KntIqKsi32fWy92bKuKYuScZ4xzkuqoiAvS1e2TwqjumfLGVlbkNtNKJGwhjS1Nq5OqpUwcKPKrSDw9MoospMgJraVmmJQkUGlsS3wbRRVWZNlNZWeTgybev2cZgKrPz0fBJ77lxjWR8ZzhwkO24Rk4HAsNeTCh0HTcKkeGgd58EVNUTD2drs9pesivPX9/X3Onj07dYwQClFKeQw5TdOpYtjCOpFzEi88NEbGGE+7DDM9Q70W8cBhwh+XaxYDLh6rHEd+V4x/k5Ui3wmhmpCVEsIc4eQYrjYEMpIguBj0kE2zvr7OwcGBV7GU/SVrVu5LWCQFJtIH4cogNPoh1THk7h/WjppxB6Yebus0TwdU/SNsBCII9g+9fjmA26EoanZ39siyMe35NjNJCrWiqksMTmIay/BQBgpdU5U1OlYkRtg6kdcfVxpqbF1QUC7CijX6tUBmluFeGuODwbHACcpi3xZuAmPEgBpqjKMPWnplmqTYoCwUfkVgseY4iqiiiFTFwcrSUGoNtSbDasLYjFQ36UcuRuNojkYZd0MtVCSZwLVWGF1TVVZiQfoCR7+MlV01GKOoXIk+qwevscm3kQt+24BqO5lFpTF1UVEWQ3Rt+fGNbp9izzT7VE3tdZiXLglZz29HxnOXFhrlkLMuTYx0+KCHAUw5RhOSEYM4Go04ODigqirG4zHHjx/3Giw7OzssLS15Hna/35+i9slxBWII9WrC4GfoPYPVUQk979BQySpAsmDFS5cVghhCEQELVwfh/QiPJd+DCVQh0EsTLw+9Zzm/pscr78VbD/8JhBIeC2B7e5tbt25NiagZYympp06dIkkS9vf3PdsohN6aSUqySgnHRbiqCPs/nPQPg5d+4e0QT1xN4efPojKhaXg2sCb9ZtPjDwbbDEZD4kgxNz9Pd27WQi8uackoW/au1rWlC+rKORPy7IByQVahdFg989pzuW3FIbcvxurNVLYEXl2JAqU1ylGsSCIrQAbKKW9pYiKSJKaV2vqnrSRxdV8l4cfJj6mIKIppRQlxaitBxQ4yotZUuqIqS6rK0pF1XaF1jfJEfWu0ZeVhlPKwU+TK8Vnn3mXFepvjJlyHp2jtlDDdfpWWwLJUc9KkacL8wiKtOAVjKGoXC2z05zPvnxdQDfvdT+STjNjnLQOOjOceetmhMZXPpIWfxXHsi1WEBgqYMoIhNTHPc693YoxNCBIRK9FRn5+f5+TJk9y6dYsnT56wuLjoz2s4HFKWJcePH/eGPvSoAY9Nh8FUpZQ/h/D8wmsXSmMoDyzHLYrCl9gTQy0TjEwoIRe+1WoxHo8pisJLJUj6f2goQ0MdZrOKYFrotcu5N1k3cs3SJxK0zrKMmZkZVlZWplZOoZZOnufkeT7lrR820ci5hvIT4XU0A8by/SNl2IGJzza9tBYPXvn/NxKVaMZXnTftV7O2GtPeQZ/RqAcqZnFpkZWlFaK0RV0U1HqSYVnVkSAWk/NSjuqHpfspE6NU7ZQhlaVH6tgmF2mXQIZx9EhtszH95KXc92L//VhZHXelrWGPogSFJopT0lZCkiaoUmGSCkopGmJVGk2kUFFCmtrSdrrW1HGEquxFGDHmwo6JQdVhgXrrqdsYg3LaMnLXndOAcd5uNGUytdH2+o3IM2i0rtC1tnBXbWME3W6X1ZVlWu2UYW9AVeTTtivsy2k3/vkBVaFp/j3bkfHcmxS/MEgZsj2a8Ix4y/LZYd8XbDiES+q65uDggE6nw5kzZzDGsLGxwZMnT7z079zcHA8fPvQKkaFYl1RBagZUw2LYIXMmLEfXNFJinGRCkGBpq9Vibm6OxcVF7/2LpzsYDIiiyCcDCWQS6rWEsQaprxoaw1DDJVSwDL1f2VegkxBrl8CyrBJkwonj2F/rwsICX/3qV7l27RpgVzHz8/OAnSSGw6GfoMNVWPhQyErosKB7aPibQfXmqu9INDPJdDRubW3fg0RXjd+HiTfH5Hu2NWSPnec+HA6tyF0MiwsLnDi+SquV2LJ2ZUFd1uRZRpGXlJXF2Y0xvtqQ1tZgGR/AM1OGh8gyVybjxMI8dW0s4CPHMVZ/xXvNSqQGItqtlJmZNrMdEbdr0UpTkiR1Rt96/KJmKZTJKFKo2FIh48QxYWKXEasmBa/F440iyQVQ/h5NXiYBYe3+KdcfFlIi6Cdnn4ztBV0byrKirCrqUltVSQXd2Vnmu/MuoJozysboyhYNt9+fGgZTfTd1fgY/HuR6/JJuakg/33E5MsZdWmjkmxzykKYX/mt6rmFQUDxfOW5VVayvr/PkyROyLKPb7fL666/Tbrfp9Xqsr68zOzs7xUDZ3d31XqwYYCmCIUZWPpOsTTmXMAAr1yCYt0wEMuGETBeppjQzM+OpgzKxiFEPJ72Q0RMGZAE/GYhhFjVGuTcSR5DvhoHqkI7ZarWmKJyh9y+/LdcrsgJSTUpiHaF3L/BLWK1KxoBMPGGuQxh8DoO6zbHThGmOVGt6Z40yaiZUezzs2Z2Cc4LAs0uUGQ+GHOwfUNcwO7/IxfMXmO8uUtc147E1OFmeU5QVuNWBeLBWidEGAbVx3HMsvVE5OAMs00Z+OnIp/0mkHMxhz8ZCHDVlXTk4wxCphJmWNehz3S6d2Q6d2VlmZtq+lJ4Jnltr3+wEYs/QBcqVlS6WAKzANMqdm63G5PB6FZHEYgdw3rv93J6/vZbY1YiNJUjfwMCV8/ptRnFNpR2kZSowmihSdGbnmOvOogyMxiWjUWljDAHsNuk782z/BoHU546LoNef146McQ+NS7jsDyVmw4BgiKeGWLu0JgYb8s6laLTAIJcuXWJlZYXhcMi9e/c8u0WSg6Io8oW1xasOvXM5rnjy0gQekaCjBDGbwdMweCxwhfyuePThtcg9ECMoRjysCCXGXrz18Xg8xdoJ7594+DBh1BgzKboh5yGxAfkn+7bbbX+d8t3d3V1f3Wo8HnPz5k20thLJi4uL/twHg8HUhBGqQYZQW3juzTEjHn/o/YfXdqQMfBNXNUFEFAfauOf40PTyQzy+cJIdDoesrW+RlzmdTpuz589x+sQJNIZRNiIfjhnnOZUuXIFq6/nWtaauNEVRUZcVlSt6gVLY6kUJIoBlsAZdSvElsYVK4ighimKvK58XFWVRUhQVpjZEESRpi9nOHLOdWebn5unMzNBKZkjj2FIiNRZyqVyQv9bouqaoalufVNuQp60UZZOnoigmdiUFtbt39tXNQip2gV0b/NTG4CTf0UaKwtjVQJzamrGI1+/olNpYqQYpv2eqCuPOSWsbsF3ozjE/O+eetz5l1oda+1KDk+UYBBb/mab8VBXa+HB/w5cZ0kcGcw8DlmHQsfmQh0yVkEXRXLYfhrdKFmWe52xvb3tY5sSJE1y+fJmNjQ22trbo9/vEcezT6AeDAaurq1OYtBjf0LsMhb86nY7Hl8fjsS9kIVWexGgDU7x2gVXkOHmeT8nzyjUJVh3yymX/MJgrvPEQkhGjKBOEnI9g4GFwU347z3NbUMF5700IRI4XRVa6YXt7m2PHjrGwsIAxhr29Peq69kU7RMpBYgwyBqSPQ4Ms9zGkl4Z9LNclk5FcjxzrSOHuavJi3Jvw7Dwz4kuwIQDvXdvxUzEajtna2WIwHnBidZWzZ85w7epX+OjmDbJsRJGPKcuKWlulRuX0WhS4IhfW801MjIoVkXGYuZIyddAyCbWqiZOYmbSFjhSx1lZLPrITfFkBdU1palp1gkETxS3aia26NN9dcCvYFmVRUNb2WSqiwhky7fRcrD2MHDxhS+PFGANJVGNsLhaV0ajaQlMAMREmxgY3E2M1YpSiqCtUDaUrImKLeqeksRU/S9z5F1WE0hqtFO0ktZ6zSaiiAmNqCq2pdE2lS7SuaCWzHD9+gvljiyhjGA/H9IcltRHNGuOhNd/Pn9enwZgwSnkoTwWDRnTen9eOjOfezHxsBkdDYx0GKSHAL5kk5oSTgniUYQbl/v6+52W3Wi1effVV2u02g8GAzc1Njh8/zvz8PMePH2dzc/MZoxhSIoUxIri61DwNRcZkhSGJPDI5SBMOeAjZCCtHOOQhBDIcDoljK1ssE0sY2A0NeZjOL/dPoBGBjET58rAAqiRyCdQiXnoINwkmnyQJT548wRirArm6usrW1pb35N944w2OHTvmr380GjEajbxipdyncIIJA7dh8Y5wAmgGeZsrtiPTpnDfSfNBViYP9qGozCFLeYvnWi83z0u2tjbY3lin1jXziwu8fOU6y8vLaANFYZklWldopUijxOHcVgrYklkkO1NkP2zd0yiNrbRuEpPELVpJi6SVMpPa1zRp2WPFERGWRmlqC/dYT9llrCYJSTuh1Z6h1W6RpC3i2NFunTZM7dgoHt9XijhKXKZtYtUmI0WsYitnEMVEkXKKkpGTOE5cndUW7TS1vxslLqDqbENkGTtJ2qaTdGi32sykM3TSNmnSZiZpkaYJrbhldXlw1aKqnKqsqJ26QGeuw4VzF1nozmOqmr2DA4ajHtQ1ug4897AfGx39bL9PAqlqarb/cjP/kTHuTcghhFFC7+uwbU0PNDT0IZQj2/M8p9frsbW1RVEUZFnG1772Nc6ePUtZlty+fZtOp0O322Vx0eKVa2trHpIJuecw0YMRgyUeuXj9IVQgwcvD9FBkspAJR6iPwNSkIfro4s2GUEYTj5bJRyAbmBSYlmsJjyPnLnIK0kQDpkmHDGEepZQv0HHq1ClWV1c5d+4c6+vrnrWzsrLir3s4HE4Z8jA3QI4ZHlvOXbbJOIFp+YrDVnxHpRn/P/e3D9bZDySYJkHV6S82AnHhdpeAU9UV25u7PHxwh/FwRJzEnD93nssXL9Jqt6jLnCLLqarasj2UxIFc/dPaVkGqKk3tg6EWA1YoiOXeY1Uonffvt1kQ31U4qqmN1T2v6oKyLKiqwq52jbGJP5HF7f3vB7TKsioszbEunUcuz73FoLXRk0CqkgQqFxB2io9KRcSJoz5GynH1BdN2MgixxdxVqibQjEtIilIH9cX2d7WpKYuKPK8pyoqqtpK+S8tLnDxzmvbMDONyzPbeFr3+kLoyLoELpno0DI42+nYyD5ipbZP2vOl/0o6UcQ+X203mRDNYJn+LJx166OLByYMtxk2gDQkwrq2ted2TkydPcu3aNeI4Znd3lwcPHrC6uupL392/f5/79+/T6/V8YDCkJorBkgLQwqkXpkyIjwvcIecuEIzAOmKoZYI5duyYZ++I1g1MOP7iScvfkpAlBlgmixCbFVaLMcbHAgSrl3OWWIGsEjqdjo8dhF57mFR0584dHzNYXV1lMBjw8OFDlFJcvXqVS5cu+fPb3t72cA1MY+hhIP2wcRKupMKVkfR/KK1wlFpI3oBJYE95cFVeGw/uFzzPFhe2E2+v3+fO3Ts83dykNprl48t85epVFheXqIxmPB4xHI3Is4yyqi1PWymLDeMmeW3ZIJUY+9pWFbKsGOHjGzSW960xDoIQL1VZNgsxkUVZLD+8rsirgrIqrHHUVj3SBnFLiqqmqi2VthD+eq0niowKu7/bLy9Lu39t67ra+2CNfa0FDFEeAsO4yURFk8CwMGQczbEWsoC9sRMWkTGUZU1ZFeg6pypL6soQJzFnz5zhxImTRFFMNhqzc9BjnGVoXVvt+OY4PGyJFrwqPucz19v27/9MAqqh8Q4fVPkMntWfkYdbPL/QkxQjIUlBYaalwBxbW1tsb297SOKdd97h9OnTjEYj/vZv/5Y4jlleXmZ5eZmFhQU+++wzbt686YN70uGdTsd77MK0ES9cfi/0IEPKY5iNKwFQoVlKADdJEm/YQ6ldERmT6zXG+IzZsGReWNM1zOYMDaAY/RD/FzpjSNNsrjYE3y/LkocPH7K9vc3KygozMzOcPHmSn/zkJ+zv7zM3N8f169dpt9soZfMFDg4Opgp7hKuIEF5qnmMzBtPcv5nEdJTaM4+j5JobCaB9rvX+nAPIvbHGqaoqhvt9bt+6x517txlnY+a6c1x78QoXz56llaaMsyHjXo/BcGRjUEVlDbQxVJWmrCpLlcxL8qIgz0vnIFXU9aTMXF1VlIV42VXgnIjOTOQhH4Ohrm3uQzYeMRqNrSRunlEWpS35VxqbLVqWnmpYVTbBypgAhtUGrSsHX9p966pGays5oN3KQ2vr3FQ+LoYLEE/ucV1rdG0DyUVRMs4K8sxec55XFEVFlpf+syy3//LMeu7G1HTnZrn60lWWlxZRaAajIf2DA4q8tKuLz8Pimn16WB/7OX96XHwZn+ULjbtS6rxS6k+UUjeVUjeUUv+t276slPp/lVKfutclt10ppf4XpdRnSqmfKqW+8cWnwTMPZ3NbGLCTv40xHosWAy7GQQyBHAcmRRxEsGp3d5eNjQ3Peb927RrvvPMOcRyzubnJp59+ysmTJ1lcXOTUqVNeyVAMs1D4QpZIWDFJoBrxpiVNP1yViICZnGcIj+R5zt7eHgcHB97TlxWIGL0QlxbYKOSdixcr+LoUBZfzF1qkXIcEZKUmq0wmcm8FjhLlSlkNPX36lPv37/uM34sXL7KxscHNmzcBWF1d5erVq/47wi6SgHQz3tIMqoYrM2BqHMj+IYUzvCe/6LHdHOfhK0JFDG18I8gq+029Nlo4+WVFycb2Nndu3+Rg74A4TXjx0ot8/auv24QyZRhnI0bDPv0sIy9KS110bA5tLI2xEnimtobWaOsdow1a15SVTYjKy4q6tAa+rrWlJEYx7SQlSRNSH0sxVGVJnheUeUGRFRRFTVWUtuyfv8QIjEYZi5unScqMaCe5wh91DaaWnAsx3nY9UTutm8pDfva8reqjDQyLyJkt7mGTksrKTmJ+oqrc6qGsyIuCosgZZhlZnpOXGXVVkMQR5y6+wKsvv8L8XIcyr+gd9BmOc4ypnREOQ6m+xw4ZHIf0cTBOhPyDkgS4//+wTAX898aY68DbwD9XSl0H/gfg3xljrgD/zv0N8H3givv3z4D/9Uv8xqFJKqEXHgbWDsOvZWCHxiGEL0IjAHjP+Pbt29y/f9+LhL322mscP36coij48MMPGQ6HLC0tEUVWLbIsy6mkGwmaNg1pWAkohF6aRkdWF+FEFsIdwmeX1UYYZxBDG8YWhMkSeujye81EH5gO5MqEkiSJX82E9y007iGu3+v12NjYIEkSZmdnuXDhAhcuXODu3buMRiNmZma4du0a3W7Xyz7s7++zs7Pji4yEFaBCpUyYnpxDwy+v4bhowjZfAMv8XMb2dHsWRDUB/npY3lD4tc9bydvjWEihKAp6+31uf3qX+48eUdUlC8uLXLv6MpfOniFSKYNej4Nej+Gg5zOZq6qm0gZTiyaLhT6q0tL/ysp6uWVVUxWaoqqsV5+XjPOSsqgoKzcBgGXRqMRnqGqtKR3dUlhrWTYiyyeQaYTNco2ihDiOSFstW1UqSa08QWQ11I0WCQBX4Lq2115Xli5Z6RpdOZ59KZ9bPZvagC2fZLVxjIayMlRlTVGU5EVOXhRkpYWPsiIny3L6oxHj4YB8nJFnNuN3fnGBt77xTS5dvghE9Hr7PH36lH6/Z++XqTFaQPQQaG8Y5uYYcPtMSUALU8ZjTM9vX2jcjTFPjTE/du/7wMfAWeAHwL90u/1L4Dfd+x8A/7ux7a+AY0qp01/id6Zw1Cbjo4m7w8R7C722poGHiWKjHFf0wMfjMdvb26yvr7O/v8/29jYXL17k/fffZ25ujs3NTX70ox+xsLDA2bNnmZ+fJ01TNjY2GA6HHsYQ4xIGQCXRSAy9wCNCX2xCIqFqZbMghaxQQhGt5j2TSa7dbk8xj4yxzByBjmSSaJauCwOj8l6Mu3jI4eoo9KJ3d3fJ89xr4V+4cIH79+/z6aefEscxS0tLXLlyhTiOvcTvYDBgf3/fJ4fJPZM+lHMLJ+bw7/AeSF+HQXi5lnAc/KLG9vOaMYJeiw6JYL3i6wXMCBME3Zh+xUEz2gVVR+MRj9bW+eTTG/QP+qStlAvnzvPSiy+z0F2gLGqGBweMBn1GwzHjcU6eFzaTVbsSbn6itN67GPuyriwmX1YOFimcx1v61Wyla8u68SWHXMBTV9S1hVLyIiPPC4oyo6xKa5DRVqYggjRpMdNq0W61aLXabsVrJdZqZ5W103DXCKPKacB43ZmKWlt4R2sbSzDh/kb2t9clXntZ2u+WRUmRFwzHIwaDEUNH4c3LnCSJuPjCZV575XU6c7NkwxFPH69z99FD9vs9O6noSRzCdvXEMYVJkNU0x4AfF757g/3suy8y738vzF0p9QLwdeCvgZPGmKfuo3XgpHt/FngUfO2x2/bcFmYThgYr+G3/GmZHwiTVXL5z2DI9hHvE+Euq9vr6OuPxmL29PdrtNh988AHXrl0jSRIfSO12uxhjOH78OPv7+9y9e5eiKLzXI1mYwtoQXjvgPX25tqqqvBFuJmBJ+T8JyAqcI5xwYb0IFCW8cxEwk21CxQz1bOTapQSevBdPWc6hKArm5+c9Lz+cBEKmjHy+vr7uNXmWlpbo9Xr85V/+Jb1ej3a7zXe+8x1eeeUVAJ9M9fjxY8/RP8wAiz6N3LPw3jUnxiY0JZ+F3vwXtX/Isf383528+jHuP3X+e5itOL1KbzzgDprR1gM9ODjgs1sfs76xjsawdPw4L1+7zvlzFzBxxMFwwM7OLr3+Af3xgHGeUZS5M961MzLWI9bGsmdqbT1Rb/Tddq2d52wsDGKMBP2UVZFUtqpTkqREUQzGZrHaPIyasiptwWkDUZzQaXeZn5tlYXGR+YV55rpztNptKzfgzsuxv23GqhJpXksn1VgDr4315MWYi/HEMZIk09RCUG6yqtyE4ILKWZkxGg/JxkOK8ZAsy6lrw/xcl1dff50LF85h6prtrS0+vXOLh2uPyEZj6tpOJFOsj3SnAAAgAElEQVQAuQ+gSz+6iK70ZQjDTI2TZ8fws1DPdPvSSUxKqS7wfwL/nTGmFz6MxhijDvv15x/vn2GXtsC0JnvT+xav+zBjGJwDMJ0W38xkVUpNFXqQYz99+pTd3V1mZ2cZDofMz8/z/e9/n42NDR4/fswPf/hDvvOd77C0tMRwOCSKIp4+feqN0+nTp/05C44ufHUxSKF2i6wclFJeEExgnJCvLVCFFMSW6wsNllJqijMv1xzGImDihUsTXF4MtwRd5bihmJdcg0yYgsmPx2MeP36MUorV1VXP///0009ZX1/HGMOVK1d46623/Kqm1+uxubnJw4cP/YomvHdNTLq5SpFrDD9rJiw9D2f/nLH4Dza2nzkXZ68nHpl9uM0zOAzgWB5qKqtl6mV6m2DmtU1oGg7GPHr0lAcP7nPuwjkWugu89OKLvPG113n65BGfPbjHcNCn3+oQKdt3ceR0jaIEUjE+Gm0imyEaXobCM7GVKxgShfdfG5QyrkxdQqwUaE2SprRSO2Z1bQ2pjTtYqYB2q0UcxSwsLrB4bBEVRVR1jak1pa7QlSGJbR1YlSiSOIFIUSmoyso6x9qet40hWDjGOINuS+NZGqjSUJXaBmDrGm0gityEUGuquiYvc8ajnNFwzGhsWT6z7RbXr7/C2298jW53ltFowMO1B3x27zY7O/sUmWXxWOZN4KFL/wtDx0wMvPE78Aw0Z5wi5xfY86n2pYy7UirFDv7/wxjzf7nNG0qp08aYp25puum2rwHng6+fc9ummjHmXwD/wh3fn3LolclDK8ZQcG2YLuAcwjYhvVBa02OXJsccDofcunWLY8eOsb297THi119/nSdPnvDkyRM+/vhjvv71rzMejz3OvLW1xfz8vPdkZeUgSVLCS5frCg14KDAWYttVVXlFRwlmimyuMGMEBpJrbbfb/h51Oh1/7SHuHp5HeA5Zlvnfkc/kfskKYDAY0Ov1PA1zPB6zu7tLr9cjz3MWFxfRWnPmzBn29/f58MMPKYqCU6dO+Umxqiru3bvH3t4eH3/8sV/xNJkuIZ4e9l84UYeTjXw3hHLC7zeP1Wz/0GM7juPpxzHw1OUBD1MSPSki9OJV4/viyMskH2yz9t160WVZsr9/wGeP7vLq/nUWugssHV/lrW++ycbmBgejAbs7O/QHB6gkZma2QztJ0Ao6zJBECbFFQTARpHFk6Y7GoJV91tppC4Mt4EEU2bqmaZvaFZnGPRNJmhKZ1BbgSGI6c7O+2DbaYCJFkiYYpUjilHarxdLKMseWjmGMoiozxuOCqC7QuiJNW75+QquVooFSYjNak8QxqlZEsUFhhcvSxEocGAxliINbvQSXMGXVLbWuKRzZICvGZNmQbJyRFWOiBK5cucqv/9qv89Kly0QYdrf2uX//IY+erDMejJzDIrRQM91/06Nlqv8Ciz+tb6OMHxNf1sB/GbaMAv434GNjzP8cfPR/A//Uvf+nwB8E2/8rZdvbwEGwxP3cFsIyMKEyhsvtw6iQ8l35WyANmKT1h15fSJmM45her0dYvLmqKkajEXNzc/zjf/yPuXbtGkopbt++zYMHDzhx4gTHjh1jaWmJubk5er0ea2tr9Ho9b3T39vaoqoq5uTn/O2EAVVYnYXZo6I3KtYqxFw8+zIqVa5SsW4F6JMAr8EyYuRrCIHJ8kTcQoyhMFjmuUpbPn2UZe3t77O3tsbm5ydOnTxkOh/5+i4H/6U9/ysHBAbOzs7z77ru88sorRFFElmUcHBzw8OFDnjx5MjVBNSG2sP8E5mrCaiF7KlzZheybkDl1WPt5je2weccsAFONA9NNGG/zmCzTxiFAaIw5BHcNoKqyKOn3xjz87A5373xGfzAkSmNOnT3Dd771Dq9df5XZ2Q55ltE72GXcG5JnBaasQUXESQSR1ZJR1kQCEpC0PPYkTenMzNLqzNCZsZXM4jQhSez3ImdiYifX22rPMNuZo92aoZ22rQSCSoidbk0SJUSRopV2aM+0SWdmaM+0iZIWSZJaTRkSWpHLPG3PuOzYlCixdV+tzkxEHCcQx0TKZrUqGRMojHbsndpQ1baik7CBysoGVbPxmNFwyKg/oj+0LBmtNSdOnua9D97jzTe/QWduhtFwyNO1R9y/d5/trV3HAqtcEXE9ZYvVVFcG3rzr02eCKmG/+/HxxdID8OU8928D/yXwkVLq79y2/xH4n4DfV0r9N8AD4HfcZ38E/AbwGTAC/usv8RvAtOqjeGMCF4hnFhryELaR1vT4Qm863EcmgqqqvHjVxx9/zLFjx9jY2GBmZoYXXniB3/3d3+X3fu/3uH37Nj/5yU+YmZmh2+16RUmtNbu7u54VkqYpy8vLHDt2zNcPBbyRT9P0mVqfwo4RD70oCubm5mi1Wj5pSgx3qGXTDJqK8R8Oh57vLtmoYRKXwDxJkrC7uwtMY/0C2UgAVLTZh8OhL3YSSvdKIPVHP/oRa2trtNtt3nvvPd5++23SNGU4HPL48WPquubBgwfARBtHoKtwJRZOgE24LvTW5T6EQVW55uak8Tnt5za2/diDqfX2dGC0GWTDB1a9MZcVSfA90X8PLb1g71k+Zu3xOj/52c9YWTlBO32Z2U6HCy+8wLvf+ja7u7vcuv0JoywjpodSEemMzU7WlbYyi7VGG1vAw2LY2mPWkYpQcUTbCXPFTkGy0pbpYfH4GKNrVNJmptUibbdptVLKonTeqIUutMPrlQGvl17XFpLQtS/lh6mtzK6/R1ivW1sgqzaGGEON816jyS2XKkq1g0zQE5Zd7dg1Va1drklGNs4YZiPy8ZiyLujMzfPW66/yrTe/wezsHOPBgCePn/Dx7bvcfbpGbyjaPQIP4z1xOwB8bz3TpvvUwY5K4WPSwSj6MtoyX2jcjTF/PnXc6fbBIfsb4J9/0XEPa/KAiwfWfDAPC5CFUE0Iz8B0AQl3bs8EF8WrLIqCR48ecf78eS5cuMDu7i5xHLOyssI/+Sf/hH/1r/4Va2tr/M3f/A2vvvoqy8vLbG5ueoExWREURcHKyor3fCUTNPSEQ3ZNaKTEQIvKokxuQlGUBzvEqsNrarVanmJ2WGxCDGAYxBSuuxhJMYp5nnsj3+v1GI1G5HnOcDik3W77Aiazs7N0Oh3u3r3L48ePSZKEr371q7z33nscO3YMrTVra2vs7Oxw+/ZtLzkwGAw8zBTCMmG/hucjhk2u6zBaZCg1Eb7/vPbzGtvPwOlq+u3kc1ekw78aK4EbxFemvhcu3QOjIWyLuq7Jspzd/X0++eQWS90FknbKhfOXaHfavPSVK7y5/g16vV3uPlqjNxiglSJpJWRzHUZZYo23MzK6rtFAVdYoZVUS8zilrSLKVBETOQVEa6RLXYOGOK7QqbXjcZIGGkIua7XSoFwwtnIFMKKE0XhI3G6hiCiyMdm4pKwyxrml6eZlAUpREVmN9dpmvBptKJWx3rkz+kZpCm33KYuKorIBU7ShrCrnrZeUVU5R2KSlPBszyjK7oi8r0laHl69e55d+6R1WllfJsjFPn6zxs08+5rN7t9jb26UYW+aPdpPIZPoVyO3zLfL0WMB/rwnR/Kf03H9urQnLiAELt8l+IXwB0wG45uQg6fJhZqhg2eLVSkGOn/70p8zPz3vDsbi4yJtvvkme5/z+7/8+u7u73Lhxg5dffnmKpSOGJ01T1tfXvdiWGCiBCCRgKvxyCcACviC2wDky8Ui2qQRVhc4pBs0Y439PjLtcl6xcRB1SgqGhRk632/VGNKyPaozh4OBgijETRRHHjh0jSRLm5+fpdDo+JqG15urVq3z/+9/3nPYHDx6Q5zkbGxusra0xHA7Z39+fWok1Pe+wv8Wrbxr5ZlA5/E446clq5hfZPs+wA0yUyu1fk/8HwVY+x3M/zKMLmCC1tiyUwXjIk7Un/Kj9EVErAhLOnjzN0soi3/zaNylGGXn5pzxdW2M8GtKZadPvd9AoIu0mz8gpLMYRptJEiaIyLdI4tZ6xTqgTTWJiQFHWJXVZojEkVQSmtmyZ1BVvj5UtvK21FTIzTiOorqirirooiFPLd7fPS06ZlRRVaROgypw8y9ARpHWCUUxK7RlNYuw2n+BkLL2zqCoKR9usqgqMNe5FUTh+e05RluR5xng0ZjTO0VXFTLvF1asv8/3vvcdXrl0lMoa9zS1u3b7FT2/d4OHaGoP9EXmZ26xXCaSawJwb4/t7qqMnH/tgexg/V26c+GFs/hN57j+vFrIkmlTH0Ks77MFuMi3CQKLs04RCwlWCMcbDKltbW/zwhz/k7bffRinlU/6/+93vkmUZf/AHf2ApZp99xpUrV4iiiOFwCEziBiI+durUKebn5z32LQZSMkHlPETLRQx2mqY+azWU9RXlRIEzwgkriqzUrnjGcg8kOUkMXXifsizzqwfxouV+7O/ve2kA0eYRymer1fLB1bW1Ne7evQvAyZMneeedd3jppZeoqorHjx8zHA55+PAhN2/epKoq+v2+j200+z8MostkGI6FZp+GYyCcBMIg69FrMs4loGos7OCErARHN7KP56O4Z6QRfHu2+PLE+BsHOeRZwf7BgIcPHtFJW3RasyRfizlz+hRnL57jnfg7jHXJn/zpn7C9u0uv34c4Ji8y66HXmkjFJGlEGreJk4gkjpkxhjJtoWJbnDqqY+o0RmGfN11VaKOYaRmrvZ4mpA6WiZQmUzHGaM+d19qm9Ou6oixr4s4MrfaIKEqoqpwit8yavCzInOdeY+EciNBC0TUQRRq01It140gby12vSoq8pNIlpjaUlTXsWZ75GFSRl+TjgiovaLdbfOUr1/gv/tE/4mtf+xqzczOM+j3u3r3Lhzdvcv/BfXa3D8hGI0+U0HoiO+DjoMqtzT7PKKupF0KfRLpe3HiZ+J/Xjoxxh2eNuDzUIcul6emJkZZ9Q7ZE6MWJ13zYxCDH2N/fJ01THj16xPLyMl/5ylcoy5IXXniBTqfDBx98wGAw4I/+6I9YX18niiLOnz/PwsIC/X5/iv+9ublJnuecPn3a66BbL6Xy5yUYe1mWU9WfwkIY8l1J98/cMlFS940xvgbsYDDwUgKAx+/l2sNYBljOuWDtIo0g3PXBYOAZPqE8wcrKil8Z9Xo9Hjx4wHg8Zmlpibfffptvf/vbxHHM/v4+m5ub3qvP85zxeOwxdukzaaLDI30SsmHCvm566+GrKFk2oZovwN3/4ZsJmQ72vY+bTQXIGvdF7lPwfxW6/oLHH7YwcZ9ZqLAERigFdx4+oNVKaHdmmOmkrC6f4Oz5s7z1xhtsrW/ww5/+mEHvgKrSdDpjYme4UbENiKZt0iSh1bLMl6LdglhRFrYgtiojYiLHW7cVirI0YUFrlIpI0oRWmmABHotzW695TFFauKSsatpJQtxK6LRt8fmyKqhLTaVLirKgdAlQGjepq4iyrjCV9hNgFCnKukaXkrxknFee22PUFbrSlGVOltt/eZ5RFBVlbn+HSHHm3Fnee+99vvaNV1nozjEeDVlbX+OTWze4c+8uO5v7DIcDK2JWVROev+9H/DlNvHi7TYaGndRVY9t03/ph7A/y/BXpkTLuYoTC4BowZQzEgIdGXWh+oaEOvddma3qIsn9RFGxvb3Py5Elu375NlmVcuXIFgBdffJFOp8Nv/MZvMBwO+Tf/5t+wsbFBnuecOnXK489aTwptDwYDHjx4gDGGU6dO+XqnYHnne3t7rKysIKwVgRCaDA/x+IXJs7Ozg7BfxIMWkTHBWmUFEWLs4UQnsJdcf3gPQwkC0dIpy5KFhQW/fW9vj4cPH5JlGcePH+cHP/gB77//PlEUsbW1xZMnTxiNRty5c4der+fpoaE0b9i3YZ82YyRN2CWc+EMsXibAMFfgF27YweLm8seXQYjEEEzRHqdJM8/8xjOHdk6MwcIzZcloPGJ3e4fPkoR2Z5bFuUVmXpnl2LFFrrz4Fd75pT32RiN++tO/Yzjok2Vj2q0WKrYSAkmVUFYVSZLSMW1bVamYgSghwlCgUFVEqiLKqgJtS0632pUfa620TZrO0FIV4yQFBVVVMMgyyixnXNUoDbStl13XNUYpJylQWVimKMgL6+HXCqgkh6XGSv5aCWFtlIV4ajtpmLomzwvGZU6RiwSxTVTKC4e156X33GtTsrqyypvfeodvfP01unNd8mzM47V1btz4hBuf3mV7c5fheExeFk56YZLkNekbMz0pT/XYpOcm+00+m9qiJMjK8weDa0fGuMtD2kz4EUPUTMeHaYgmDAiGXl/TQDRZFOJJimcrhks8+Pn5ebrdLltbW1y8eJGFhQV+8zd/k9FoxF/8xV942uPp06f974k3XlUVw+GQ+/fvs7W1xfnz5+l2u14QTNLvT5w4QZ7ndLtdH5SV44Tp9uPxmF6vR6/X87roIaw0Go28cQ+57rOzs/7ehXVaw/sqr6JvL+cHTOHwm5ublGXpa9AuLS3xwQcf8Pbbb3tWjVBD79y5w+7uLlVV0ev1POYvK7GQDRNO6M1JWfovNORhfzbHkYyHI9PkgUQ9441Jm0JbBEt/BlN/dn8V/E3jvXG/rbWhpqIoYDgYsb21w93WXRbm5zk2v0Bn9gqzi/O89tXXOeiP6e3uc/uzjxmOhoxHGWmrTStNMW3tqklrcgXtNGM0Tl1CkCs0ohRFHE8CoyoiHscUhVVIVHFE2kowJiJOW0Qoq0SZjxjlFp6JooRONEOatmyBjTim1Fajva4s/1wb6/sbY6i0FQUr6xqlDUSGorQ3uqwshl+UVlohL3LG2Zg8s8bdfl5TloUNqBZWTqHWNbNz87x2/TV+6a03WFpcIMuGPH26zmef3OLj27d4uv6EQX9AmduAq9aiI/M8LsyhA2Sq354JqMq2STDmixAZ4AgZd3lYwyBkaLSbgbRwnya22kx4CbXOm55c6MXL741GI1+N6caNG3S7XX+O586dY35+nt/5nd8hiiL+9E//1GuSr66u0u12fXWkoii8MNbBwQH9ft/rzYihk3T88+fPe0hHjOT8/LwPigrmLpTEc+fOeTle8VrlPkgQVXD90IMPmTHC5wc8vi5qmb1ez9+bsJLTcDhke3ubsizpdrt897vf5dvf/jatVsvz2Pv9vmfPNJkx8vtynqF33fTEwyDrlCfUgNXkvfR7c9I/Gk1N/n+IxQ5HpXjs4SpcTZn/5mfNIJyaMvDKaLSOoK4pyoLhsM/61ha3PvmE5ZUl5pcWOHP2AsvHl/mlb30LtKbVTvjo5s/oH/So6wp0G1/UWdnVyHA8QkWRlb5VorIIcZx4WELFVvK3yDLqWhMhdNeImXZKq9WmRlOUzlMHZtKYbmeGxYUF5udmrS57XZNjXA1Ve4WxUrbYiGJSqMMYtIYIO+FoXVOVFUVeUBQlWTFmnI3JssJpxdfUlaYqrcxvVZagDPPdY3zjta/xvV99nxfOnqEqcp5urPPJxze5+fEt7q2tsbvXZzzOnFyB1ZHxY7nRxU17f1ggfOKlT1gy4fdVJBO+msz8z2lHxriH3lkILRzm2U2xBhrG+bAHP/TmwonhsEAcWC9WKiitrKzw4Ycfetii3+9z6dIlZmdn+a3f+i3iOObf//t/T7/f9xODUCGNsVrv4/HY498h/7yqKg4ODtjd3fXcdtFo11r7oiByPwQ2kqAP4AXQJOjZxLSHw6GHU+T+yUQwHA79NYu2+mAwQGvtM12FAST7iEjY3Nwcb7/9Nt/5zneYm5tjMBhw9+5dX+jk7t27xHHMcDj0xwxjIRIkbgbNm5Bc6MXLOAhx9mYgPnwvE/vRaPLQT4Kn0DTM8oiHy/AwwPasD3+oV6+mP5948LbIxTgriA56rMXrfPTRz1ha7BInMxw/scKJUyv8ynu/zOLxY7T+nw4/+tEPba5FUVh4RH5TG0bGMnLG4xZS2SiKbPHqKIpRyiYvJUrRz0eMi4xRnpO226RGUdZWOz7LHeZeQxQr4lbK3GyXubkOcSulzgqyumCYZRSVhT+00Wjsz2psJSvjSvOhrCKlLRBin5ssHzHOLG997GSOtRynNlYorChBwcLCEm+98Sbf/7Vf48XLF4lieLq2zief3OTHN37Go4cP2d/cYzwaUTihNPNFEOCkIyBkvrhREY4L+WxqNaYsTjflBHz+rwFHyLgf1uRBDuVzw+1h8CzEWkODLTc8nCzALeecgQkNQMifDr3XH/7wh1y9epXTp09TliWXL1+m0+nwgx/8gJmZGf7kT/6EwWDAwcEBcRx7jrfouYtXLL8d0hqjyGrVdLtdLl++7GGQfr8/ZdxHoxH9fp+5uTlGoxHD4ZA0Tb0IWKj1Hscxa2trRFE0lfwkhUb29vZ48uQJKysrUyJloS4OTFZSBwcHbG1tobVmaWmJd999l1/5lV+h2+2yv7/P7du3GQwGPH36lEePHhFFEbu7uz7Q3PS8w76SvpSgbYinHyYvAUz1f1O64DAv/xfegsinZ8cYNanVEf5/Op7qjTNMiAZh8M1jsIHDM3F+5BgOf69rUDAeZ+zu7XPn7l3iNKXICl6+/hqnTp1mbn6ON7/+DTpJm+XuPH/9dz9iZ3ObsipdQpAhra0iY1VXjKPMeZKurmlk+zZSijhO0cowMztDd2uDNE4p8owkjejv9dg+2GG/d8BwkFl8fibB1IbKlIyyEbUxjEcjdvb3GPaHZGUJGorSEgWsQnBkeeyuOIZSNsCLgnGeMR5ljMYjRllGnlm6Y1lU6EpiUAZdV6BgeekYb37jLX71Vz/ghUtnQWmePH3Cxzdv8qMbN7n74BF7m9uMxvY4Vj/GUW/dPbb33gVNVNCJrrOmFXqkj6aD6qEvP2HdTLz1yaT/+eP7SBh3eUDDv+U19O5CI970YkI2jfz9PNwdmIJqQg8ynDx6vR5xHHP8+HFu3ryJMZZSeO/ePS5fvszi4iK//du/zfHjx/nDP/xDNjY22N7e9jovUgtVvF/x7oWTLr/X6XQ4ODjgzp07LC8vc+LECR8DMMaws7PD1tYWs7OzPvt1Y2ODEydOTBlHMfK9Xs/r5AidU5QlR6ORl13o9/s+xhEW6RDufVVVbG1t0ev1UEp5w/7+++/T6XQYDAae8vjo0SMePXqE1laCQX6jyWyR9yGcFsoiSDC8uboKDbVARc0xIPuK1k4zIPsLa1N4jHv9e5zWMwFVFR7j2Qd8QqiwX5x83/gU+1E2ZmtrB2NuUY3H7OwfcP3KVzn/wgusLB/jq699laWlRU5dPM9f/sf/yMPHjxj0+4yrMXVbo01KXE+sjb3X0aS2ahTRShJA026ldFpt6kKzs7tLEsf0h2O2t3cYDm1iEpFCo0nTEfv7feeFR/SHYwb9HqNxTqkrlExSuL6NAKNcrVVjr17hCoJbbz3Lc7IioypspSdbjMPKA2itiWKbv/HGG9/ie+/9CmfPniAvRqyvb/PJp7f42Y2PefTwMXu7+wxHY4vh1/XUuJzOPQ0Me9gXMgJM0KdTgfOGXpAyQfxUNoZs+c8fREfCuIsXLUE7mC7eEXpwzYe86aGHfzcf6jAbNNQJl5VBCBuEHrOwU44dO8ZHH32EUorz58+zvr7OuXPnmJmZ4dd+7dfodrv863/9r7l//z67u7t0Oh0WFhYAvFcsSUqiLjkej6e89DzP6ff79Ho9n1wk6fsAKysrfvUgRUZE2leyR4UuKcyWra0tfw7j8RiYrFAEAxeDKBmyWmuGwyE7OzsMh0NarRanTp3i3Xff5fXXX/ce+4MHD7wm/vr6OnVds7u7y8HBge87aXJPZRIJ+7cpGBeOhSYsFx43pEmG8FqItR8F7z1wrA8NqH6Z79s3watqbHuGlzHBcO0JOO9Qa2oDhSlsVaVaUxaaXr/PzuY21/de4dr16xw/fpKLL1xgrrvA2ZUT/Olf/gU/+emP2dvdI8tG1GXbwX0KlMY4AbRISU0GRe36MYpj0IbewZ7NUo1i8rJmOBoxHI7JijHKQFmmaGMDoLsHMVorX/2ozEtqbNarl84lQjnjbgtjaGf6bA3VqrLUx7KwhbarwipmCmURUxO32pw6eYpvfv1b/PJ3v8X5M2fIiiGP19b47NZnfHLnMx48fMqBy/0oHMtG1y6AOtWmYythdTzvtzdWZofJC/hXM/ne1Hg45F2zHQnjLgHA0NOW7eFSHSYPerg8b6auNwOnTWxWvtvE2sXIhDQ6gSWE9XHixAlu3LjhkxX29vY4d+4cJ0+e5Jvf/CaLi4v88R//MR9++CGDwYDxeMzi4iLdbpeZmRnPJ5+dnfWl7wRfF8kDrW3KvmjViBaMePp7e3tef77VajE/P+9xdwmaJkniKYjiFed5ThRFdLtdz7tXSk2pQiql2NvbI8syD6mkacrly5d59913ee2117zYmag8rq2t8fDhQ4yxyU/9fv/Q/gjjKWFfwiR5Ss4jXFXJ/gKfCUtIJogQkgnHxvOkB37ezY6xL1ce7RkvPdzmg60NeuUzBgLE2IOVMfDbjNU3VxqKChiDMVsUVc5omLE/HDEaj7l+/RXOnDrN6vISb33zLVaXVlnozvHXP/kbNjd2qIqcuipRUUwk1aYV1EoRaQvLWI0Vm9CT5RlJnBDFkTt7Ra0r6rKiqivAEEcJozwm7vchUpi6pqw0VWXZLgZXSCRcDSrxmsFYrQEE5tC1Y9dUUlnKeu61LjFApzPP9Zeu8svvv8fXX7vG0uIC4+GIhw/W+OjGR9y5e4/1rR12D/bJxiNXtNs5RALFEN77aVP8TBg8yC71z4WS4uTThMlp7F2uU/peJon/DDz3JnwSGvEmLANMGegm7BIet4mzh0qBMnk0A7RNr1D2EW/02LFj3Llzh+FwyIULFzwj5vjx41y9epXFxUWWl5c9VVLKyYlsbrfbJU1TVlZWvMcuMIIIagE+EDs3N+eN2cbGBg8ePPAwjzGGlZUVdnd3/TmG5x/eg07HqvbNz8+TJImXFhaMXRKN+v2+DyQUAQoAABzuSURBVMxKHODXf/3XuXr1Klpr+v0+jx8/ZnNzk729PZ4+fUpZlj7m0JRlkH9ijOV6Q29brqV53mFfN7H4cNKX/eSY8htHLku1YYHDgKp3ys30Z+G2yWI8OFC4v6wQPFQw+f/UEQzYmqM1hbFQTV0byrwkK2yN09FwyMvXr3PuzEXm52Z56eUXieKYufkuf/3jH/P04UOGwz5lUYGSQGoEKsJETrirjqwueqkZjzOU+89gdd6lUpPYa/t9EctyJRRrG7jFiBSDwFGWnWO/OzGtUYBXa619Gb6qtpx3rSsUEYvHlnnrG2/y/ge/zKuvXGcmadEf7PP4yRNufXKLW3fusP5knX5/yHg0dlWmJmyzZv/ZDdOAydQkLTfeTPeIgDlqcnXPjA+YTGDKHC481mzqKCxZoygyUukoxM6bRjmkyIU3ODT0gh2HBhymDbVkWIbbm1h96MWHx5DU+1OnTnkM+tKlS5w6dYqlpSXOnj3rz/Wjjz7i3/7bf8snn3zCaDTyEr+zs7MsLi767NAw+SrLMh/hF7YM2OIaUipvMBj4lYOIjO3s7HgFTQnISnA0TVO/TQLIgoULFTPPc0ajkb8vaZpy4cIF77Gvrq6SJAnr6+s8evSIwWDA5uYm9+7d87i8fD9kKclvNWGv5iTaXJnJ9nAlBZOas9K/zf4Mjyf/3Hn9QoD3OI6NxAf8OHoe3t6A5Q/9rNEOZdE8s00573+C5/oVrVJEsSKJUlqthM5ch+Vji5w9fY6XLl/kK1euc/niJRaWF1CVYnd7iw8//pgf/fhvuXH7E7bX18mLgiiy1ZZUnEyqIjmDryLvdk4uQ4lxtq/2NCNv7HAl6rSlwkx9T8kqSIkxDIpZuGNgFGGJQK1rICaJY1ZPrPLuO9/hV7/3Pc6eP00SRQx7fe49uMfPbt7go48/5sHaOge7u64UYGmF03TtJsXD4T7TBNGZdl6/qPlLUM/WNZgcQ8a2Ii9y6qo6dDQdGc89xEhD4xyKgIUefLjkbrIjQsZLOFmEQbzQsIQJUOEEEvKm5fclCFlVFaurq96T3d3d5eLFi547fu7cOd544w1Onz7Nn/3Zn/Hnf/7n7O3teY0WgTvCYKvovEhQU2AWweaF8x4GGEXeVxgxkoIfTl5Cu5RjSqar0CGF426M8RPPlStXeOONN7h06RLz8/M+OenRo0cMh0M2NzfZ3Nz0nv5gMJgy3gKvNBkuoZEO2U7yeUh9DCeJsL9D9o3EDcL95HeayW+/2DYJuqE+x0oTbD7sYzWBZzy8ExoMb7RhaoKQKj7eILrdxet1PPjaVDb5Zwg7GurKkI0zBv0Be/s7vHj5JZaXjnNsdZlvf+tbvPDCBc78+Mf85Z//GQ8f3mM0zqiqmlhry5iJY2IsVXGqdqh44AGcMnGg7CVNnnU9+ds77BbyEThGri1SkiTmJAi0sca4rqlqDVHE/GyHy5de5Jff+TbvfPstVk6uUpYF21s7PLi/xu17n3Hr0095+nST0X7PZbKKYbfSx5N77SAW6V5//926wRinJxMaaIKVihjwz+nTQ4MzoTbN8yeLI2HcYfJghkyWptEPje7U0ijwrOU7oVEWPB/w3mtoOOBZLfkQ6jnsuKPRiPX1dZaWluh2uzx48ID9/X0uXbrE0tISeZ5z9uxZzpw5w/e+9z1OnjzJo0eP+NnPfsbW1panTErt0XBCE4hEMlTDCUkConJdgC/IIYyYpncuxl147YLHh5z4VqtFt9vl3LlzfP3rX+fatWs+GNzv93ny5Anb29v0+33u37/vs1cPDg58cDjUsQlhrsP6WLaH+Pth8ZPDCm6EEgbST01WlYyXo8Nzt1b12YX337OFGU5iRPB/2l3CZf/U82+Cl8k5GANKG7TSmMrBJMbqtld1yTjP2O/32N3b58qly5w9f5alpRVeunyJpaVjnF4+zo//7q/5+NPP2N3eZZSPqCVZLk6IY7n2CV0wNOjPXGIIw8rOcheVo5G6jB5FIPeNwqDDOwLG0iU7nTlOrB7ntauv8EvffpuXv3KF2fkZ+r19nj59wt37D7h/7xFrT9d4ur7Dwf4BWZHbBCVhxWjTuIeHwS7Kwy8ehTGBnTY8a5KfKbV4GJjm78BkZROeyGF7HgVYRillpCTdYSwI+fswPnMTL5fX5v5yjMPwfWmHwQTSQqPT9O7TNOX06dN0u13iOObUqVNeUGx1dZUTJ04wOzvLaDTiyZMn/If/8B+4ceMGvV7Pe+NRFHnN93BCk8/EG6+qytdcFXgilA82ZlJFSYy8GHfZVzRsxNAvLy9z7tz/1965xdaVnXX8t/a5+pxj+9xsx7nUSabTDiNAMKqqSkU8o74UXhA80EpUlAcQIPFSygtSn0BQCSSENAikVoJWSCBRIRAChMQLLbRVr+lcnGQyiXPzJbbPxfa5LR72/tb+9vK244gkdk72p8mc7bVv67b/37f+61vfusiVK1e4ePEiKysrbrHT1tYWGxsbjk+/desW3W6X4XDI1taWUxiajklzX9VtIHWnr9G0m/+cNPdWTddoheL3DfFEGh0xdH3W4tMyRn28YZkUQCjL2/8sE33Xn3rziHvNPxsvzb008e5Y5ZjAYEwY9TFfyDNTKlGemaFWm2WpWWf54kWurlzmAyuXOH/uAvNz84yGE3a2t3nn+tt8751r/PDa29y7dZtevxdy1BETHkQKLqRrYmoqzEIE/TY+Dn8V0Bnj6J7ABLH1LArDgjVjLJDPBZRLM8zNNrm4dI4PvnKF13/yJ7i6colWu85kPGHr0QbXb9zknevXef/2Gusbj+js7rLb68WTp9EuaM7A0HrD1Z36cW1ErGWPgdi0uDO63WR0kqR2pH/DwWBwJC1zZsBdlvhHf7tf31rX6Xr1qr8C0re2fWCWybY0Tkv/6jxoWkeuNybeSanRaNBsNh29srKywsrKCpVKhVarRbVapVKpcHBwwM2bN7lx4wbXrl3jzp07zsNEv9e3OmVUI+6OpVIJY4wLCyyWsfDQWiFJCAJxdSyXyzSbTRYWFlhZWeG1115jbm7O7RS1vb3N+vq6o1zW1tZYX193XjsS511A1p/cFPE5c62wdZ0LwPvnRPyRk64P3Uc0FSTpe3t7ZwbcId1ifQwGxNdJ39R3eo9L4/T9fh4/i8SbQ9pGFiAF5HIFCvlws5mZygyNuXma7RbnFxb4wJUVXrn8IS6cX6Y2W4GJYbezy7s33uM73/0WP3rrLe7cW2Ov22N/ENKBIU2eC8E5MG5yUFvpExtubC2WrIk4dGNimjQXhJu7WgsEkTkcuWIWSzMsLy7z6oc/zI//2Gt86JVXOLfUolKtwGRCp7/DvbV7vLN6nbdWb3L37hq7nQ6dTi/aeSzcBHs8jnaDirx9HLuUsNat/2ekNO3xbaoUhPOekQZwFnyyXzhlrBRhyLmnzyedGXDX+41qt7zofAKkNegJGGovCQ00GvS1ZXfUkFA/16cI5Hk+uGjlUSwWWVxcpFarkc/naTQaLC0t0Wq1KJVK7nh+fp6DgwNWV1d56623eO+999xEpYQfMMa4SVexvIWqGQwGlEolR60Ui0X6/b4LEiYbeeu4OtVqlUajQavVotVqcenSJZaWlqjVam43KdkjtdfrMRwO2dzcdP70wtFLOAGpDxFNl6WNqPTqYQ3Akq55eF2vWgFIum4/yYMobF9B98M422cG3B33fayoL/wIccadMTGqyLmUfu6OPR4+yXvE+QtXmQYEuVB5F/J5SqUilUqZWnWedrvOyvkPcOWDr7By8RKtZotKZQY7ge2tHdbu3uXdWze4d/c2N26/z8b9+/T64TZ0k0kY7RHhsYUKgkO+4/H3ZkI2JhBwDydsc0GOQqFEpVRibn6eq5dX+Mgbb/DBV1+l3axTKhWjVdg9Nje3uHn7fW7deI+bd9a4v7FBJzJU9g8OGI9GjEYhTz+x8SjS4YCqrgQ2OLA+jKeuHRzvYl27WZUWllWa0mu/cPATK70XBdyDILDiCSKFSvNdh8OLlNIsPd8CljRIbuHnUy36N8rXIc8PP+SBPFPnR9wXxbe9VqvRbDY5d+6cA/5Go0GhUHCRIHd3dx3A37p1i7t377Kzs5NYZTkej13MGmvjzTcmk4mz5o0JNxgRBZnP52m329Trdebn51laWmJpacmtgBXfewltsLGxQa/Xc7snaS8Y8dIBEiMYXU+6fuT9/lyGHw3Sd3eVXz1C89tYiw/0/vvOhreM9MWT+bofJe7j9xP1ny4tVhChQRhRGMZgbAiU8cRf+GChFEILO6ZQAiMurDlyQZ5ioUClOkOzXmdhoc3F85dYPrfM8qVl2vUFZsol8rmA0WDEbn+X99+/zzvvXuP6nTXW766x09mm299jGIUMsBOwRD7qVmKyR2AexhgI85DPkcsXKOZy5PI5ZsplmvUmi60W5xaXWLl4mZWVZRYvnqOYLzAaH9Dt7rG9s8P9B5vcvnuHe/fv8uDBBrvbu/S7HfYGg3BrvPGY8XjiwvZOhOtBUTCKMwvbIenpHvbbw7SaugJlsieul3ZLpCUUsHF0FTZcyzV4EWgZiUvuW8WQ1Iby0co18iEft4JRW4NpQ38NKJpi0EpCc7h6glPuSVMixhiazSbVapUgCKhWq1y9etVtU5fP52m1WtTrdbdgyRjD5uYmW1tb3Lx5k9XVVRcyQLanE68XAXAJc2Ctpdlsug0/2u2226xb0uv1uvOnl71jdVjevb09Hj586Db9Ho1GLhKlHrVI3WkvJD1ZKnWulWla3Jg0UPdXmWoKLi1GkLzDnwiXe6NRzOmCe7QsH9Jpk/+PCL2SwHjPmtcYcRw9JLRAfMpEK0Bx1FdggnA3pnyBcrlErVphfq5Gs9GivbjApcXzLCwu0Gq3mavNUiwXGQ/HdHZ2ebixxd3793nw8C43b7/P+oOH7O51GOwNIo57wHgcUXiEq1yDINyar1wpUa83aNTr1CpzzM1WWGi2WV5cojpbo1YpM1utUZwpYrH0el0ePtpg8/4mDze32Hq0w8bmFjvdXTqdHgcHe6EzwnDofOFDj5iJG0Wo/3lUio9Lh9PSzom3lBVi/QnwV+ZbZJQDLwjnLtamBgB1HiAVFOS8v1Rd8+/JyYjkPZqqkfs1XytgokMR65gSvjKB5AYguVzOAbeEAxbPlHa7zdzcHKVSiUqlQqVSYXFx0YUTEJfFfr/P1tYWq6urLvqjeLsMBgPK5bJ71vLyMqVSiUajkQhaViqVnIfL5uamU4hCs8gKWVmdKpZ6yEEOEzSKxKv3h/36Gt0emlMXENai6953S/W9aSaTeF9dn57x8yJyNiz3iC+GJAqnEbMpaTEFE+OB8azzxPVp9KOyzt01crcAf6AUhbj6GaIl/6FiCgIDQUA+CK3nUjEM3VudqVKbrbJYb1BvL7C00GZxYYGl5SWqM7MUCznyQbTH6mTExvo699cfsLGxQ7e3y85en421+3T2uoyHllwhoFyuUApyNNttlhttLly+QLPRolTMUSzPUMrlyZcCDvYPGByEfPnewZBOd4d7D9ZZX99gc3OLzm6PTrdHZ6/H4GA/UiQjRhNNwViwQhPZGMiF24/qSazsBLjLuOixlnt0bDV1mLTS/XOJNOdu8xQmVI0xl4AvA0tR/t601v6pMeYPgF8D1qNLP2+t/efont8DPgOMgd+y1v7rY95hxbrVYCsfufh+p1kaacN2bS3Ked+yhniyT5/Txxq4BaQkkqT8nUZJ+MpHXBbz+TwzMzPUajVmZ2cdLdJqtZibm6NSqdBoNAiCcFcnoXaE5llbW3MeOblczkVdrNfrLpxAoVBw0SLz+Tz9ft/FVRc3yskkjv/e7/fpdrtsbm66nYzkWFbK+iCh61fTaP7Es1ZyvnKFw2Duu7hqkNfujz4X77eZHpnB0eD+PPr2SSdUTyoazmNgTrHc4zIeOufyISMIbeF7ykBc8hOGEUBgQv+XnCGfC92Li/kChVKemZkqtVqV+dosjfk5lpYvcK61wPz8LLPzdRaadarzNXLkGI2GjEeQKwQMxxM6nS67nR2GownlUoFcUKSQD7dQrFWqlCsFApNjMhqGe6gOhnT3e2xvbfNoa4tHO126uz22drfZ2NpiZ2ebTq/P3l6f4XAQxnEfDxmPI0vdhi6Ok8ifXvhwDaBGpx1hnSOWvmobdUpZ7lHaUW1JMhqkuHU6jh6iSejw7MHB0YuYTgLuy8CytfbbxphZ4FvAzwO/CHSttX/sXf868BXgo8B54N+BD1lrj1xNYoyxs7OzqR+tHl7HlXXYc0Ku0da/UBi+n7TjEs3hfTn9UYBPG2gw01aor2S0aLpCRgHCx0toAeHK2+025XLZKQFRDDLh3Gg0mJmZcas+ta+6THTKvqfGGBcDXmgcCQ8gq1JlcRSEm3H3ej16vV5i9ORPSotHjW4TqWcNtP7yf61sdZ1KHUndau8XnQe/zRIWqDc60+ePAfdn3rdzuZzVrqdxnk82Ivct9riMyRglj7vxsAXv8bmSFgXhis+hAE2sRiP/OddJCT2Qy+XIF3IUiwWKhXBV9fzcPI3aHNVKlfl66B680G4zO1slXyhSK1eYnZulUp0BG2AZhxO5QQCEoXwHo32GgwmD8ZCDvT67vT7dbod+74CdnR12tnfY2d6m2w/jtff39+juddnf32P/YMB4NAy3wYscE8aTiYvBHveZuIrECTNJvRzm3OO2PHoVquuLUbtx6PpYHSQMlkNamUPv3j8G3B+7iMlaew+4Fx13jDE/Ai4cc8snga9aaw+Am8aYVcKP4b+Pe4//8abRM1rSgFSDhlSQT5/o8/r5cu1RNE6aAtBeOXq0oT9if1IWQrfER48ekcvlEpEihf8Wd0VN40iEyd3dXcrlMp1Ox/mzS+x1vfephCeQbflkoZO1NoySF4G9bPAh8dxFKeo61T7o4/E4sbOTrmddhxAvNvIlTYH77SEg7wO9HomlrXD1332clfy8+jZEQJh4eZx+HMY7i87K/db9IhOkKe9yXVcOPGA69HZ3XWQpuvxFMKdAxRBGXgwt0cjGnIwxJoh2GQs42B8QmH26hS472zusl0oUS2XKpSKVmQqzs3OUSmVmZkrUyjXm6jXm5xoExTyFXEAhVyRXMEzGll5/j36vQ6+/z2CwR6e7T7fXiYyRfcbjQbgBx8EBo+GA4XDMcBT262Hkzih9KcGpi7X9GEtc10fyWptsG/85CXBW8WOsUW0YjwyS7cuhoZiRsyexCnjCFarGmMvATwPfAD4O/KYx5lPAN4HftdY+Ivw4vq5uu8PxHwyAs7JFfOD1+VQNKCp/qcCdUo5Dk6oaGCRN3qXv8QHJF59aOso1UKiNXq9Hv993QcmEgpENtuVZEqZAwF5Wt0qMdglrIBOrYr1PJhMH/hJmQG91J4Au+dRupj495tMdafWqLXC/jeS83266nqUO/Xf5z/K9bHwqSOflJPLM+rbru0l4t95v4oRJSXJcb+yfoSMJJvsXMTcc4bfViG/1kw+/K7bqZeJPrZ50oBda8XYCxk6icobeLuLhYgLDcBSwf3BAr993dGI+H1AsFClETgUShqNcKmOCgHwQ7ptqgtCrpx+B9mAYhu0djMdhRMrxJPJyEVfhcTgxaicwIY5I6X2H1tqwHKqeonlOVdeHgdqiGsKvwRSeRSgU1ybutKZ+vL6glH6SpPf7RUpH8eTE4G6MqQF/D/yOtXbXGPMXwBeit3wB+BPgV5/geZ8FPpvIrleh+kP26RltjR8FFNp6FovUB4S01ZQ+PyoWoo4xrt/ljww08At9o6/1yyP0RRAEzoNFwDufz1Mul92HIRtya0UoFImE7hULRax3mSD1t+DzPYeARB1psJRQBnpE5YO1VgYystEWt9SNHh0dpSR1Xep21qCuQwP779dK5nHyLPu265tRDJLjwsqAwgZvRG51bBZ1bapVqekW8OKbJC338D1JasD6lyVf6I4t1m1zZ2y41Z2BcDbChCGF47YJyEVhgcN473nCZg0XTJkgIGdykctogDFR3U1gzITJaMzYjglxOdx8W9wnbRQgLGHYSV3pY1dXus7iKtN1n2yPpBWfKm7zFJu4Kv4rfkOyGq3XHfyRU3SdKAxpnMdj+8nA3RhTIOz8f2Ot/QcAa+0Ddf4vgX+K/lwDLqnbL0ZpCbHWvgm8Gd1v/YlNzXsnucokzyrHviuk9y4HEr5VqF3rfJDQkub94fOovmumzqu2Zn3O2V+4o0MNGGNcUC69gYm8T1vi4vcuItSED77yDD+Imq5XXSe6LJInH7TlPVo5+HWQpkTTwNcf6Rzl6y6K5qi4Miex2p91387l827kLwGlErNz8V3IdeFBcrB+JKigYFlAXaGzjU5YhAqLF1DFdRbnR+iFxNuOeLVY9eHIIAQkDYJG8fbGjBmhApgxSPYNcKEF4glFt24nAnENwPGK1hg64wwks2wj5Xh0HR4a1CB5P1wXvuKNryXxnmTfjuFeK5K0vOpfeVdilJF449HyWHA3YQv8FfAja+0XVfpyxFkC/ALwg+j4a8DfGmO+SDjp9CrwP497T5pnylEgqc/7ikCuk1+xYnWcFQ1YvpKQ5/iWps6H/ltASPO/aQrmKKDzFZoGNR/0xfL2wxLIs4QL1/WiJ3wlJHAaKKcpIG3Za+5bv8Mvj9yj3+uX06etfOWdpjg04Mvz9a/uH3Lv4wD+efTtwBgK+UL04KRVLt9pIi1p5MV5VZObzto0yWfpGCWH0kzaY5NHTkFYZdk60RAak0P+ExL3HZPXRJn0pVbgL/x7Aok3GVW2ACMRB+I6DHzgFDXh8xuHKli2gU0BzePL7eYl0ArTulS/TQ+luZxp4E/mDe966euDiHJNk5NY7h8HfgX4vjHmO1Ha54FfNsb8VPS694Bfjwr1Q2PM3wHXgBHwG/YYb4JIut1u9+0T5GVapA1snHYmnpOchbKuHJH+zPv2aDTqdrqdrG9Pp5yFsh7Vt8/MIqZvWms/ctr5eF7yMpX3ZSprmrxs5X+ZynvWy3p2NpnMJJNMMsnkqUkG7plkkkkmUyhnBdzfPO0MPGd5mcr7MpU1TV628r9M5T3TZT0TnHsmmWSSSSZPV86K5Z5JJplkkslTlFMHd2PMzxlj3jbGrBpjPnfa+XkaYoz5a2PMQ2PMD1Ra0xjzb8aYd6PfRpRujDF/FpX/e8aYN04v508uxphLxpj/NMZcM8b80Bjz21H6VJb3SWTa+nbWr1+w8iaW7D7nf0AOuA5cBYrAd4HXTzNPT6lcPwu8AfxApf0R8Lno+HPAH0bHnwD+hXCJwseAb5x2/p+wrMvAG9HxLPAO8Pq0lvcJ6mXq+nbWr1+sfn3alvtHgVVr7Q1r7QD4KmHkvRdarLX/BWx5yZ8EvhQdf4kwtKykf9mG8nWgbsJQtC+EWGvvWWu/HR13AImsOJXlfQKZur6d9esXq1+fNrhfAG6rv08UQfIFlSUbL2m/T7hBBExRHRhjLhNHVpz68j5GXpZyTn07v6j9+rTB/aUUG47jpspNyXiRFfW5aSxvJodlGtv5Re7Xpw3uJ4qyNyXyQIZp0e/DKP2FrwOTElmRKS7vCeVlKefUtvOL3q9PG9z/F3jVGHPFGFMEfokw8t40yteAT0fHnwb+UaV/Kppt/xiwo4Z9Z16OiqzIlJb3CeRl6dtT2c5T0a9Pe0aXcJb5HULPgt8/7fw8pTJ9hXD7tiEh9/YZoAX8B/Au4d6bzehaA/x5VP7vAx857fw/YVl/hnBo+j3gO9G/T0xreZ+wbqaqb2f9+sXq19kK1UwyySSTKZTTpmUyySSTTDJ5BpKBeyaZZJLJFEoG7plkkkkmUygZuGeSSSaZTKFk4J5JJplkMoWSgXsmmWSSyRRKBu6ZZJJJJlMoGbhnkkkmmUyh/B8te/yzOZ7ZwAAAAABJRU5ErkJggg==\n"},"metadata":{"needs_background":"light"}}],"source":["from keras.models import load_model\n","autoencoder_model = load_model(\"autoencoder_for_skull_images.h5\", compile=False)\n","       \n","import random\n","num=random.randint(0, len(img_array)-1)\n","test_img = np.expand_dims(img_array[num], axis=0)\n","pred = autoencoder_model.predict(test_img)\n","\n","plt.subplot(1,2,1)\n","plt.imshow(test_img[0])\n","plt.title('Original')\n","plt.subplot(1,2,2)\n","plt.imshow(pred[0].reshape(SIZE,SIZE,3))\n","plt.title('Reconstructed')\n","plt.show()\n"]},{"cell_type":"markdown","source":["Getting just the encoder part"],"metadata":{"id":"WMxm1sNOR0V9"}},{"cell_type":"code","source":["input_shape = (256, 256, 3)\n","input_img = Input(shape=input_shape)\n","\n","encoder = build_encoder(input_img)\n","encoder_model = Model(input_img, encoder)\n","print(encoder_model.summary())"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"18cN97JQP_6d","executionInfo":{"status":"ok","timestamp":1665077724186,"user_tz":-360,"elapsed":646,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"733fedcf-7d26-4c58-d26b-28e05e0468a1"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Model: \"model_3\"\n","_________________________________________________________________\n"," Layer (type)                Output Shape              Param #   \n","=================================================================\n"," input_4 (InputLayer)        [(None, 256, 256, 3)]     0         \n","                                                                 \n"," conv2d_57 (Conv2D)          (None, 256, 256, 64)      1792      \n","                                                                 \n"," batch_normalization_54 (Bat  (None, 256, 256, 64)     256       \n"," chNormalization)                                                \n","                                                                 \n"," activation_54 (Activation)  (None, 256, 256, 64)      0         \n","                                                                 \n"," conv2d_58 (Conv2D)          (None, 256, 256, 64)      36928     \n","                                                                 \n"," batch_normalization_55 (Bat  (None, 256, 256, 64)     256       \n"," chNormalization)                                                \n","                                                                 \n"," activation_55 (Activation)  (None, 256, 256, 64)      0         \n","                                                                 \n"," max_pooling2d_12 (MaxPoolin  (None, 128, 128, 64)     0         \n"," g2D)                                                            \n","                                                                 \n"," conv2d_59 (Conv2D)          (None, 128, 128, 128)     73856     \n","                                                                 \n"," batch_normalization_56 (Bat  (None, 128, 128, 128)    512       \n"," chNormalization)                                                \n","                                                                 \n"," activation_56 (Activation)  (None, 128, 128, 128)     0         \n","                                                                 \n"," conv2d_60 (Conv2D)          (None, 128, 128, 128)     147584    \n","                                                                 \n"," batch_normalization_57 (Bat  (None, 128, 128, 128)    512       \n"," chNormalization)                                                \n","                                                                 \n"," activation_57 (Activation)  (None, 128, 128, 128)     0         \n","                                                                 \n"," max_pooling2d_13 (MaxPoolin  (None, 64, 64, 128)      0         \n"," g2D)                                                            \n","                                                                 \n"," conv2d_61 (Conv2D)          (None, 64, 64, 256)       295168    \n","                                                                 \n"," batch_normalization_58 (Bat  (None, 64, 64, 256)      1024      \n"," chNormalization)                                                \n","                                                                 \n"," activation_58 (Activation)  (None, 64, 64, 256)       0         \n","                                                                 \n"," conv2d_62 (Conv2D)          (None, 64, 64, 256)       590080    \n","                                                                 \n"," batch_normalization_59 (Bat  (None, 64, 64, 256)      1024      \n"," chNormalization)                                                \n","                                                                 \n"," activation_59 (Activation)  (None, 64, 64, 256)       0         \n","                                                                 \n"," max_pooling2d_14 (MaxPoolin  (None, 32, 32, 256)      0         \n"," g2D)                                                            \n","                                                                 \n"," conv2d_63 (Conv2D)          (None, 32, 32, 512)       1180160   \n","                                                                 \n"," batch_normalization_60 (Bat  (None, 32, 32, 512)      2048      \n"," chNormalization)                                                \n","                                                                 \n"," activation_60 (Activation)  (None, 32, 32, 512)       0         \n","                                                                 \n"," conv2d_64 (Conv2D)          (None, 32, 32, 512)       2359808   \n","                                                                 \n"," batch_normalization_61 (Bat  (None, 32, 32, 512)      2048      \n"," chNormalization)                                                \n","                                                                 \n"," activation_61 (Activation)  (None, 32, 32, 512)       0         \n","                                                                 \n"," max_pooling2d_15 (MaxPoolin  (None, 16, 16, 512)      0         \n"," g2D)                                                            \n","                                                                 \n"," conv2d_65 (Conv2D)          (None, 16, 16, 1024)      4719616   \n","                                                                 \n"," batch_normalization_62 (Bat  (None, 16, 16, 1024)     4096      \n"," chNormalization)                                                \n","                                                                 \n"," activation_62 (Activation)  (None, 16, 16, 1024)      0         \n","                                                                 \n"," conv2d_66 (Conv2D)          (None, 16, 16, 1024)      9438208   \n","                                                                 \n"," batch_normalization_63 (Bat  (None, 16, 16, 1024)     4096      \n"," chNormalization)                                                \n","                                                                 \n"," activation_63 (Activation)  (None, 16, 16, 1024)      0         \n","                                                                 \n","=================================================================\n","Total params: 18,859,072\n","Trainable params: 18,851,136\n","Non-trainable params: 7,936\n","_________________________________________________________________\n","None\n"]}]},{"cell_type":"markdown","source":["Saving the encoder weights"],"metadata":{"id":"-rS9_O3fR64R"}},{"cell_type":"code","source":["num_encoder_layers = len(encoder_model.layers) #35 layers in our encoder. \n","\n","#Get weights for the 35 layers from trained autoencoder model and assign to our new encoder model \n","for l1, l2 in zip(encoder_model.layers[:35], autoencoder_model.layers[0:35]):\n","    l1.set_weights(l2.get_weights())\n","\n","#Verify if the weights are the same between autoencoder and encoder only models. \n","autoencoder_weights = autoencoder_model.get_weights()[0][1]\n","encoder_weights = encoder_model.get_weights()[0][1]\n","\n","#Save encoder weights for future comparison\n","np.save('pretrained_encoder-weights.npy', encoder_weights )"],"metadata":{"id":"dzwh8UsGQGCx"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["Saving the pretrained model for unet"],"metadata":{"id":"0YvUjM_LSIKg"}},{"cell_type":"code","source":["#Now let us define a Unet with same encoder part as out autoencoder. \n","#Then load weights from the original autoencoder for the first 35 layers (encoder)\n","input_shape = (256, 256, 3)\n","unet_model = build_unet(input_shape)\n","\n","#Print layer names for each model to verify the layers....\n","#First 35 layers should be the same in both models. \n","unet_layer_names=[]\n","for layer in unet_model.layers:\n","    unet_layer_names.append(layer.name)\n","\n","autoencoder_layer_names = []\n","for layer in autoencoder_model.layers:\n","    autoencoder_layer_names.append(layer.name)\n","    \n","#Make sure the first 35 layers are the same. Remember that the exct names of the layers will be different.\n","###########\n","\n","#Set weights to encoder part of the U-net (first 35 layers)\n","for l1, l2 in zip(unet_model.layers[:35], autoencoder_model.layers[0:35]):\n","    l1.set_weights(l2.get_weights())"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"fgW2LM74QJwK","executionInfo":{"status":"ok","timestamp":1665077800119,"user_tz":-360,"elapsed":1199,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"ba5fc3b3-6858-4d72-fd9f-b47f5025db30"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Model: \"U-Net\"\n","__________________________________________________________________________________________________\n"," Layer (type)                   Output Shape         Param #     Connected to                     \n","==================================================================================================\n"," input_5 (InputLayer)           [(None, 256, 256, 3  0           []                               \n","                                )]                                                                \n","                                                                                                  \n"," conv2d_67 (Conv2D)             (None, 256, 256, 64  1792        ['input_5[0][0]']                \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_64 (BatchN  (None, 256, 256, 64  256        ['conv2d_67[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_64 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_64[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_68 (Conv2D)             (None, 256, 256, 64  36928       ['activation_64[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_65 (BatchN  (None, 256, 256, 64  256        ['conv2d_68[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_65 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_65[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," max_pooling2d_16 (MaxPooling2D  (None, 128, 128, 64  0          ['activation_65[0][0]']          \n"," )                              )                                                                 \n","                                                                                                  \n"," conv2d_69 (Conv2D)             (None, 128, 128, 12  73856       ['max_pooling2d_16[0][0]']       \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_66 (BatchN  (None, 128, 128, 12  512        ['conv2d_69[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_66 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_66[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," conv2d_70 (Conv2D)             (None, 128, 128, 12  147584      ['activation_66[0][0]']          \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_67 (BatchN  (None, 128, 128, 12  512        ['conv2d_70[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_67 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_67[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," max_pooling2d_17 (MaxPooling2D  (None, 64, 64, 128)  0          ['activation_67[0][0]']          \n"," )                                                                                                \n","                                                                                                  \n"," conv2d_71 (Conv2D)             (None, 64, 64, 256)  295168      ['max_pooling2d_17[0][0]']       \n","                                                                                                  \n"," batch_normalization_68 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_71[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_68 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_68[0][0]'] \n","                                                                                                  \n"," conv2d_72 (Conv2D)             (None, 64, 64, 256)  590080      ['activation_68[0][0]']          \n","                                                                                                  \n"," batch_normalization_69 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_72[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_69 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_69[0][0]'] \n","                                                                                                  \n"," max_pooling2d_18 (MaxPooling2D  (None, 32, 32, 256)  0          ['activation_69[0][0]']          \n"," )                                                                                                \n","                                                                                                  \n"," conv2d_73 (Conv2D)             (None, 32, 32, 512)  1180160     ['max_pooling2d_18[0][0]']       \n","                                                                                                  \n"," batch_normalization_70 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_73[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_70 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_70[0][0]'] \n","                                                                                                  \n"," conv2d_74 (Conv2D)             (None, 32, 32, 512)  2359808     ['activation_70[0][0]']          \n","                                                                                                  \n"," batch_normalization_71 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_74[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_71 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_71[0][0]'] \n","                                                                                                  \n"," max_pooling2d_19 (MaxPooling2D  (None, 16, 16, 512)  0          ['activation_71[0][0]']          \n"," )                                                                                                \n","                                                                                                  \n"," conv2d_75 (Conv2D)             (None, 16, 16, 1024  4719616     ['max_pooling2d_19[0][0]']       \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_72 (BatchN  (None, 16, 16, 1024  4096       ['conv2d_75[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_72 (Activation)     (None, 16, 16, 1024  0           ['batch_normalization_72[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_76 (Conv2D)             (None, 16, 16, 1024  9438208     ['activation_72[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_73 (BatchN  (None, 16, 16, 1024  4096       ['conv2d_76[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_73 (Activation)     (None, 16, 16, 1024  0           ['batch_normalization_73[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_transpose_12 (Conv2DTra  (None, 32, 32, 512)  2097664    ['activation_73[0][0]']          \n"," nspose)                                                                                          \n","                                                                                                  \n"," concatenate (Concatenate)      (None, 32, 32, 1024  0           ['conv2d_transpose_12[0][0]',    \n","                                )                                 'activation_71[0][0]']          \n","                                                                                                  \n"," conv2d_77 (Conv2D)             (None, 32, 32, 512)  4719104     ['concatenate[0][0]']            \n","                                                                                                  \n"," batch_normalization_74 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_77[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_74 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_74[0][0]'] \n","                                                                                                  \n"," conv2d_78 (Conv2D)             (None, 32, 32, 512)  2359808     ['activation_74[0][0]']          \n","                                                                                                  \n"," batch_normalization_75 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_78[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_75 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_75[0][0]'] \n","                                                                                                  \n"," conv2d_transpose_13 (Conv2DTra  (None, 64, 64, 256)  524544     ['activation_75[0][0]']          \n"," nspose)                                                                                          \n","                                                                                                  \n"," concatenate_1 (Concatenate)    (None, 64, 64, 512)  0           ['conv2d_transpose_13[0][0]',    \n","                                                                  'activation_69[0][0]']          \n","                                                                                                  \n"," conv2d_79 (Conv2D)             (None, 64, 64, 256)  1179904     ['concatenate_1[0][0]']          \n","                                                                                                  \n"," batch_normalization_76 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_79[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_76 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_76[0][0]'] \n","                                                                                                  \n"," conv2d_80 (Conv2D)             (None, 64, 64, 256)  590080      ['activation_76[0][0]']          \n","                                                                                                  \n"," batch_normalization_77 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_80[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_77 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_77[0][0]'] \n","                                                                                                  \n"," conv2d_transpose_14 (Conv2DTra  (None, 128, 128, 12  131200     ['activation_77[0][0]']          \n"," nspose)                        8)                                                                \n","                                                                                                  \n"," concatenate_2 (Concatenate)    (None, 128, 128, 25  0           ['conv2d_transpose_14[0][0]',    \n","                                6)                                'activation_67[0][0]']          \n","                                                                                                  \n"," conv2d_81 (Conv2D)             (None, 128, 128, 12  295040      ['concatenate_2[0][0]']          \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_78 (BatchN  (None, 128, 128, 12  512        ['conv2d_81[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_78 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_78[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," conv2d_82 (Conv2D)             (None, 128, 128, 12  147584      ['activation_78[0][0]']          \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_79 (BatchN  (None, 128, 128, 12  512        ['conv2d_82[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_79 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_79[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," conv2d_transpose_15 (Conv2DTra  (None, 256, 256, 64  32832      ['activation_79[0][0]']          \n"," nspose)                        )                                                                 \n","                                                                                                  \n"," concatenate_3 (Concatenate)    (None, 256, 256, 12  0           ['conv2d_transpose_15[0][0]',    \n","                                8)                                'activation_65[0][0]']          \n","                                                                                                  \n"," conv2d_83 (Conv2D)             (None, 256, 256, 64  73792       ['concatenate_3[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_80 (BatchN  (None, 256, 256, 64  256        ['conv2d_83[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_80 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_80[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_84 (Conv2D)             (None, 256, 256, 64  36928       ['activation_80[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_81 (BatchN  (None, 256, 256, 64  256        ['conv2d_84[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_81 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_81[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_85 (Conv2D)             (None, 256, 256, 1)  65          ['activation_81[0][0]']          \n","                                                                                                  \n","==================================================================================================\n","Total params: 31,055,297\n","Trainable params: 31,043,521\n","Non-trainable params: 11,776\n","__________________________________________________________________________________________________\n","None\n"]}]},{"cell_type":"code","source":["!pip install segmentation-models"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"dI9Q8jWJYx5z","executionInfo":{"status":"ok","timestamp":1665129358625,"user_tz":-360,"elapsed":3466,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"32d05d00-be79-4e30-8089-e0e682c1f522"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n","Collecting segmentation-models\n","  Downloading segmentation_models-1.0.1-py3-none-any.whl (33 kB)\n","Collecting efficientnet==1.0.0\n","  Downloading efficientnet-1.0.0-py3-none-any.whl (17 kB)\n","Collecting keras-applications<=1.0.8,>=1.0.7\n","  Downloading Keras_Applications-1.0.8-py3-none-any.whl (50 kB)\n","\u001b[K     |████████████████████████████████| 50 kB 4.7 MB/s \n","\u001b[?25hCollecting image-classifiers==1.0.0\n","  Downloading image_classifiers-1.0.0-py3-none-any.whl (19 kB)\n","Requirement already satisfied: scikit-image in /usr/local/lib/python3.7/dist-packages (from efficientnet==1.0.0->segmentation-models) (0.18.3)\n","Requirement already satisfied: h5py in /usr/local/lib/python3.7/dist-packages (from keras-applications<=1.0.8,>=1.0.7->segmentation-models) (3.1.0)\n","Requirement already satisfied: numpy>=1.9.1 in /usr/local/lib/python3.7/dist-packages (from keras-applications<=1.0.8,>=1.0.7->segmentation-models) (1.21.6)\n","Requirement already satisfied: cached-property in /usr/local/lib/python3.7/dist-packages (from h5py->keras-applications<=1.0.8,>=1.0.7->segmentation-models) (1.5.2)\n","Requirement already satisfied: matplotlib!=3.0.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-image->efficientnet==1.0.0->segmentation-models) (3.2.2)\n","Requirement already satisfied: tifffile>=2019.7.26 in /usr/local/lib/python3.7/dist-packages (from scikit-image->efficientnet==1.0.0->segmentation-models) (2021.11.2)\n","Requirement already satisfied: scipy>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from scikit-image->efficientnet==1.0.0->segmentation-models) (1.7.3)\n","Requirement already satisfied: imageio>=2.3.0 in /usr/local/lib/python3.7/dist-packages (from scikit-image->efficientnet==1.0.0->segmentation-models) (2.9.0)\n","Requirement already satisfied: networkx>=2.0 in /usr/local/lib/python3.7/dist-packages (from scikit-image->efficientnet==1.0.0->segmentation-models) (2.6.3)\n","Requirement already satisfied: pillow!=7.1.0,!=7.1.1,>=4.3.0 in /usr/local/lib/python3.7/dist-packages (from scikit-image->efficientnet==1.0.0->segmentation-models) (7.1.2)\n","Requirement already satisfied: PyWavelets>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from scikit-image->efficientnet==1.0.0->segmentation-models) (1.3.0)\n","Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0->segmentation-models) (0.11.0)\n","Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0->segmentation-models) (2.8.2)\n","Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0->segmentation-models) (1.4.4)\n","Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0->segmentation-models) (3.0.9)\n","Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0->segmentation-models) (4.1.1)\n","Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0->segmentation-models) (1.15.0)\n","Installing collected packages: keras-applications, image-classifiers, efficientnet, segmentation-models\n","Successfully installed efficientnet-1.0.0 image-classifiers-1.0.0 keras-applications-1.0.8 segmentation-models-1.0.1\n"]}]},{"cell_type":"code","source":["from tensorflow.keras.optimizers import Adam\n","import segmentation_models as sm\n","unet_model.compile('Adam', loss=sm.losses.categorical_focal_jaccard_loss, metrics=[sm.metrics.iou_score])\n","#unet_model.compile(optimizer=Adam(lr = 1e-3), loss='binary_crossentropy', metrics=['accuracy'])\n","unet_model.summary()\n","print(unet_model.output_shape)\n","\n","unet_model.save('unet_model_weights.h5')"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"GPo7XWz-QZvW","executionInfo":{"status":"ok","timestamp":1665078095287,"user_tz":-360,"elapsed":1203,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"34c0d9bd-2a8f-41b0-85b4-d80d2d386fe1"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Segmentation Models: using `keras` framework.\n","Model: \"U-Net\"\n","__________________________________________________________________________________________________\n"," Layer (type)                   Output Shape         Param #     Connected to                     \n","==================================================================================================\n"," input_5 (InputLayer)           [(None, 256, 256, 3  0           []                               \n","                                )]                                                                \n","                                                                                                  \n"," conv2d_67 (Conv2D)             (None, 256, 256, 64  1792        ['input_5[0][0]']                \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_64 (BatchN  (None, 256, 256, 64  256        ['conv2d_67[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_64 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_64[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_68 (Conv2D)             (None, 256, 256, 64  36928       ['activation_64[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_65 (BatchN  (None, 256, 256, 64  256        ['conv2d_68[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_65 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_65[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," max_pooling2d_16 (MaxPooling2D  (None, 128, 128, 64  0          ['activation_65[0][0]']          \n"," )                              )                                                                 \n","                                                                                                  \n"," conv2d_69 (Conv2D)             (None, 128, 128, 12  73856       ['max_pooling2d_16[0][0]']       \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_66 (BatchN  (None, 128, 128, 12  512        ['conv2d_69[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_66 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_66[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," conv2d_70 (Conv2D)             (None, 128, 128, 12  147584      ['activation_66[0][0]']          \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_67 (BatchN  (None, 128, 128, 12  512        ['conv2d_70[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_67 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_67[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," max_pooling2d_17 (MaxPooling2D  (None, 64, 64, 128)  0          ['activation_67[0][0]']          \n"," )                                                                                                \n","                                                                                                  \n"," conv2d_71 (Conv2D)             (None, 64, 64, 256)  295168      ['max_pooling2d_17[0][0]']       \n","                                                                                                  \n"," batch_normalization_68 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_71[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_68 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_68[0][0]'] \n","                                                                                                  \n"," conv2d_72 (Conv2D)             (None, 64, 64, 256)  590080      ['activation_68[0][0]']          \n","                                                                                                  \n"," batch_normalization_69 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_72[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_69 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_69[0][0]'] \n","                                                                                                  \n"," max_pooling2d_18 (MaxPooling2D  (None, 32, 32, 256)  0          ['activation_69[0][0]']          \n"," )                                                                                                \n","                                                                                                  \n"," conv2d_73 (Conv2D)             (None, 32, 32, 512)  1180160     ['max_pooling2d_18[0][0]']       \n","                                                                                                  \n"," batch_normalization_70 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_73[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_70 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_70[0][0]'] \n","                                                                                                  \n"," conv2d_74 (Conv2D)             (None, 32, 32, 512)  2359808     ['activation_70[0][0]']          \n","                                                                                                  \n"," batch_normalization_71 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_74[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_71 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_71[0][0]'] \n","                                                                                                  \n"," max_pooling2d_19 (MaxPooling2D  (None, 16, 16, 512)  0          ['activation_71[0][0]']          \n"," )                                                                                                \n","                                                                                                  \n"," conv2d_75 (Conv2D)             (None, 16, 16, 1024  4719616     ['max_pooling2d_19[0][0]']       \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_72 (BatchN  (None, 16, 16, 1024  4096       ['conv2d_75[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_72 (Activation)     (None, 16, 16, 1024  0           ['batch_normalization_72[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_76 (Conv2D)             (None, 16, 16, 1024  9438208     ['activation_72[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_73 (BatchN  (None, 16, 16, 1024  4096       ['conv2d_76[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_73 (Activation)     (None, 16, 16, 1024  0           ['batch_normalization_73[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_transpose_12 (Conv2DTra  (None, 32, 32, 512)  2097664    ['activation_73[0][0]']          \n"," nspose)                                                                                          \n","                                                                                                  \n"," concatenate (Concatenate)      (None, 32, 32, 1024  0           ['conv2d_transpose_12[0][0]',    \n","                                )                                 'activation_71[0][0]']          \n","                                                                                                  \n"," conv2d_77 (Conv2D)             (None, 32, 32, 512)  4719104     ['concatenate[0][0]']            \n","                                                                                                  \n"," batch_normalization_74 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_77[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_74 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_74[0][0]'] \n","                                                                                                  \n"," conv2d_78 (Conv2D)             (None, 32, 32, 512)  2359808     ['activation_74[0][0]']          \n","                                                                                                  \n"," batch_normalization_75 (BatchN  (None, 32, 32, 512)  2048       ['conv2d_78[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_75 (Activation)     (None, 32, 32, 512)  0           ['batch_normalization_75[0][0]'] \n","                                                                                                  \n"," conv2d_transpose_13 (Conv2DTra  (None, 64, 64, 256)  524544     ['activation_75[0][0]']          \n"," nspose)                                                                                          \n","                                                                                                  \n"," concatenate_1 (Concatenate)    (None, 64, 64, 512)  0           ['conv2d_transpose_13[0][0]',    \n","                                                                  'activation_69[0][0]']          \n","                                                                                                  \n"," conv2d_79 (Conv2D)             (None, 64, 64, 256)  1179904     ['concatenate_1[0][0]']          \n","                                                                                                  \n"," batch_normalization_76 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_79[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_76 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_76[0][0]'] \n","                                                                                                  \n"," conv2d_80 (Conv2D)             (None, 64, 64, 256)  590080      ['activation_76[0][0]']          \n","                                                                                                  \n"," batch_normalization_77 (BatchN  (None, 64, 64, 256)  1024       ['conv2d_80[0][0]']              \n"," ormalization)                                                                                    \n","                                                                                                  \n"," activation_77 (Activation)     (None, 64, 64, 256)  0           ['batch_normalization_77[0][0]'] \n","                                                                                                  \n"," conv2d_transpose_14 (Conv2DTra  (None, 128, 128, 12  131200     ['activation_77[0][0]']          \n"," nspose)                        8)                                                                \n","                                                                                                  \n"," concatenate_2 (Concatenate)    (None, 128, 128, 25  0           ['conv2d_transpose_14[0][0]',    \n","                                6)                                'activation_67[0][0]']          \n","                                                                                                  \n"," conv2d_81 (Conv2D)             (None, 128, 128, 12  295040      ['concatenate_2[0][0]']          \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_78 (BatchN  (None, 128, 128, 12  512        ['conv2d_81[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_78 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_78[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," conv2d_82 (Conv2D)             (None, 128, 128, 12  147584      ['activation_78[0][0]']          \n","                                8)                                                                \n","                                                                                                  \n"," batch_normalization_79 (BatchN  (None, 128, 128, 12  512        ['conv2d_82[0][0]']              \n"," ormalization)                  8)                                                                \n","                                                                                                  \n"," activation_79 (Activation)     (None, 128, 128, 12  0           ['batch_normalization_79[0][0]'] \n","                                8)                                                                \n","                                                                                                  \n"," conv2d_transpose_15 (Conv2DTra  (None, 256, 256, 64  32832      ['activation_79[0][0]']          \n"," nspose)                        )                                                                 \n","                                                                                                  \n"," concatenate_3 (Concatenate)    (None, 256, 256, 12  0           ['conv2d_transpose_15[0][0]',    \n","                                8)                                'activation_65[0][0]']          \n","                                                                                                  \n"," conv2d_83 (Conv2D)             (None, 256, 256, 64  73792       ['concatenate_3[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_80 (BatchN  (None, 256, 256, 64  256        ['conv2d_83[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_80 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_80[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_84 (Conv2D)             (None, 256, 256, 64  36928       ['activation_80[0][0]']          \n","                                )                                                                 \n","                                                                                                  \n"," batch_normalization_81 (BatchN  (None, 256, 256, 64  256        ['conv2d_84[0][0]']              \n"," ormalization)                  )                                                                 \n","                                                                                                  \n"," activation_81 (Activation)     (None, 256, 256, 64  0           ['batch_normalization_81[0][0]'] \n","                                )                                                                 \n","                                                                                                  \n"," conv2d_85 (Conv2D)             (None, 256, 256, 1)  65          ['activation_81[0][0]']          \n","                                                                                                  \n","==================================================================================================\n","Total params: 31,055,297\n","Trainable params: 31,043,521\n","Non-trainable params: 11,776\n","__________________________________________________________________________________________________\n","(None, 256, 256, 1)\n"]}]},{"cell_type":"markdown","source":["Declaring the metrics functions"],"metadata":{"id":"3-N9gzAzmO0W"}},{"cell_type":"code","source":["import numpy as np\n","import tensorflow as tf\n","from tensorflow.keras import backend as K\n","\n","def iou(y_true, y_pred):\n","    def f(y_true, y_pred):\n","        intersection = (y_true * y_pred).sum()\n","        union = y_true.sum() + y_pred.sum() - intersection\n","        x = (intersection + 1e-15) / (union + 1e-15)\n","        x = x.astype(np.float32)\n","        return x\n","    return tf.numpy_function(f, [y_true, y_pred], tf.float32)\n","\n","smooth = 1e-15\n","def dice_coef(y_true, y_pred):\n","    y_true = tf.keras.layers.Flatten()(y_true)\n","    y_pred = tf.keras.layers.Flatten()(y_pred)\n","    intersection = tf.reduce_sum(y_true * y_pred)\n","    return (2. * intersection + smooth) / (tf.reduce_sum(y_true) + tf.reduce_sum(y_pred) + smooth)\n","\n","def dice_loss(y_true, y_pred):\n","    return 1.0 - dice_coef(y_true, y_pred)"],"metadata":{"id":"93v0I1t-Vdb0"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["The functions needed for training unet"],"metadata":{"id":"QNr5SYGqmTkQ"}},{"cell_type":"code","source":["import os\n","os.environ[\"TF_CPP_MIN_LOG_LEVEL\"] = \"2\"\n","import numpy as np\n","import cv2\n","from glob import glob\n","from sklearn.model_selection import train_test_split\n","import tensorflow as tf\n","from tensorflow.keras.callbacks import ModelCheckpoint, CSVLogger, ReduceLROnPlateau\n","from tensorflow.keras.optimizers import Adam\n","from tensorflow.keras.metrics import Recall, Precision\n"," \n","\n","\"\"\" Global parameters \"\"\"\n","H = 256\n","W = 256\n","\n","def create_dir(path):\n","    \"\"\" Create a directory. \"\"\"\n","    if not os.path.exists(path):\n","        os.makedirs(path)\n","\n","def load_data(path, split=0.2):\n","    images = sorted(glob(os.path.join(path,\"Images\", \"*.png\")))\n","    masks = sorted(glob(os.path.join(path,\"Masks\", \"*.png\")))\n","   \n","\n","    split_size = int(len(images) * split)          # we are calculating the split size here\n","\n","    train_x, valid_x = train_test_split(images, test_size=split_size, random_state=42)  # training data and validation data will be selected randomly and the randomness seed is 42\n","    train_y , valid_y  = train_test_split(masks, test_size=split_size, random_state=42)\n","    \n","\n","     \n","\n","    return (train_x, train_y), (valid_x, valid_y) \n","\n","def read_image(path):\n","    x = cv2.imread(path, cv2.IMREAD_COLOR)\n","    x = cv2.resize(x, (W, H))\n","    x = x/255.0\n","    x = x.astype(np.float32)\n","    return x\n","\n","def read_mask(path):\n","    x= cv2.imread(path, cv2.IMREAD_GRAYSCALE)  \n","    x = cv2.resize(x, (W, H))\n","    x = x/np.max(x)\n","    x = x > 0.5\n","    x = x.astype(np.float32)\n","    x = np.expand_dims(x, axis=-1)\n","    return x\n","\n","def tf_parse(x,y):\n","    def _parse(x,y):\n","        x = x.decode()\n","        y = y.decode()\n","       \n","\n","        x = read_image(x)\n","        y = read_mask(y)\n","        return x, y\n","\n","    x, y = tf.numpy_function(_parse, [x, y], [tf.float32, tf.float32])\n","    x.set_shape([H, W, 3])\n","    y.set_shape([H, W, 1])\n","    return x, y\n","\n","def tf_dataset(X, Y, batch=8):\n","    dataset = tf.data.Dataset.from_tensor_slices((X, Y))\n","    dataset = dataset.shuffle(buffer_size=200)\n","    dataset = dataset.map(tf_parse)\n","    dataset = dataset.batch(batch)\n","    dataset = dataset.prefetch(4)\n","    return dataset\n"],"metadata":{"id":"S6awGD87iPtI"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["Train UNET for the data without pre processing"],"metadata":{"id":"MYw0_sq9mYKu"}},{"cell_type":"code","source":["\n","if __name__ == \"__main__\":\n","    \"\"\" Seeding \"\"\"\n","    np.random.seed(42)\n","    tf.random.set_seed(42)\n","\n","    \"\"\" Directory for storing files \"\"\"\n","    create_dir(\"files\")\n","\n","    \"\"\" Hyperparameters \"\"\"\n","    batch_size =4\n","    lr = 1e-5\n","    num_epochs = 50\n","    model_path = os.path.join(\"files\", \"Unet_model_skull.h5\")\n","    csv_path = os.path.join(\"files\", \"Skull_data.csv\")\n","\n","    \"\"\" Dataset \"\"\"\n","    dataset_path = \"/content/drive/MyDrive/Colab Notebooks/Skull data\"\n","    (train_x, train_y), (valid_x, valid_y)  = load_data(dataset_path)\n","\n","    print(f\"Train: {len(train_x)} - {len(train_y)}\")\n","    print(f\"Valid: {len(valid_x)} - {len(valid_y)}\")\n","   \n","\n","    train_dataset = tf_dataset(train_x, train_y, batch=batch_size)\n","    valid_dataset = tf_dataset(valid_x, valid_y, batch=batch_size)\n","\n","    \"\"\" Model \"\"\"\n","    model = build_unet((H, W, 3))\n","    metrics = [dice_coef, iou, Recall(), Precision()]\n","    #model.compile(loss=dice_loss, optimizer=Adam(lr), metrics=metrics)\n","    # model.compile(loss=dice_loss, optimizer=Adam(lr), metrics=[\"accuracy\"])\n","    model.compile(loss=dice_loss, optimizer=Adam(lr), metrics=[\"accuracy\",metrics])\n","\n","    callbacks = [\n","        ModelCheckpoint(model_path, verbose=1, save_best_only=True),\n","        ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, min_lr=1e-7, verbose=1),\n","        CSVLogger(csv_path)\n","    ]\n","\n","    model.fit(\n","        train_dataset,\n","        epochs=num_epochs,\n","        validation_data=valid_dataset,\n","        callbacks=callbacks\n","    )"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"1uL5vvhkVzli","executionInfo":{"status":"ok","timestamp":1665128658277,"user_tz":-360,"elapsed":497115,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"a207fc54-bacb-4e26-bb48-d60879ef8d8e"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Train: 120 - 120\n","Valid: 30 - 30\n","Epoch 1/50\n","30/30 [==============================] - ETA: 0s - loss: 0.7417 - accuracy: 0.7440 - dice_coef: 0.2583 - iou: 0.1489 - recall_1: 0.6585 - precision_1: 0.2435\n","Epoch 1: val_loss improved from inf to 0.81148, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 14s 394ms/step - loss: 0.7417 - accuracy: 0.7440 - dice_coef: 0.2583 - iou: 0.1489 - recall_1: 0.6585 - precision_1: 0.2435 - val_loss: 0.8115 - val_accuracy: 0.8827 - val_dice_coef: 0.1891 - val_iou: 0.1045 - val_recall_1: 7.2797e-04 - val_precision_1: 0.8750 - lr: 1.0000e-05\n","Epoch 2/50\n","30/30 [==============================] - ETA: 0s - loss: 0.6664 - accuracy: 0.8093 - dice_coef: 0.3336 - iou: 0.2008 - recall_1: 0.8226 - precision_1: 0.3395\n","Epoch 2: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 282ms/step - loss: 0.6664 - accuracy: 0.8093 - dice_coef: 0.3336 - iou: 0.2008 - recall_1: 0.8226 - precision_1: 0.3395 - val_loss: 0.8163 - val_accuracy: 0.8826 - val_dice_coef: 0.1835 - val_iou: 0.1012 - val_recall_1: 0.0000e+00 - val_precision_1: 0.0000e+00 - lr: 1.0000e-05\n","Epoch 3/50\n","30/30 [==============================] - ETA: 0s - loss: 0.6011 - accuracy: 0.8385 - dice_coef: 0.3989 - iou: 0.2501 - recall_1: 0.9452 - precision_1: 0.3944\n","Epoch 3: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 281ms/step - loss: 0.6011 - accuracy: 0.8385 - dice_coef: 0.3989 - iou: 0.2501 - recall_1: 0.9452 - precision_1: 0.3944 - val_loss: 0.8217 - val_accuracy: 0.8826 - val_dice_coef: 0.1775 - val_iou: 0.0974 - val_recall_1: 0.0000e+00 - val_precision_1: 0.0000e+00 - lr: 1.0000e-05\n","Epoch 4/50\n","30/30 [==============================] - ETA: 0s - loss: 0.5499 - accuracy: 0.8592 - dice_coef: 0.4501 - iou: 0.2916 - recall_1: 0.9732 - precision_1: 0.4308\n","Epoch 4: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 286ms/step - loss: 0.5499 - accuracy: 0.8592 - dice_coef: 0.4501 - iou: 0.2916 - recall_1: 0.9732 - precision_1: 0.4308 - val_loss: 0.8296 - val_accuracy: 0.8826 - val_dice_coef: 0.1721 - val_iou: 0.0943 - val_recall_1: 0.0000e+00 - val_precision_1: 0.0000e+00 - lr: 1.0000e-05\n","Epoch 5/50\n","30/30 [==============================] - ETA: 0s - loss: 0.5074 - accuracy: 0.8849 - dice_coef: 0.4926 - iou: 0.3278 - recall_1: 0.9780 - precision_1: 0.4820\n","Epoch 5: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 292ms/step - loss: 0.5074 - accuracy: 0.8849 - dice_coef: 0.4926 - iou: 0.3278 - recall_1: 0.9780 - precision_1: 0.4820 - val_loss: 0.8379 - val_accuracy: 0.8826 - val_dice_coef: 0.1616 - val_iou: 0.0879 - val_recall_1: 0.0000e+00 - val_precision_1: 0.0000e+00 - lr: 1.0000e-05\n","Epoch 6/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4715 - accuracy: 0.9076 - dice_coef: 0.5285 - iou: 0.3602 - recall_1: 0.9729 - precision_1: 0.5383\n","Epoch 6: val_loss did not improve from 0.81148\n","\n","Epoch 6: ReduceLROnPlateau reducing learning rate to 9.999999747378752e-07.\n","30/30 [==============================] - 9s 297ms/step - loss: 0.4715 - accuracy: 0.9076 - dice_coef: 0.5285 - iou: 0.3602 - recall_1: 0.9729 - precision_1: 0.5383 - val_loss: 0.8477 - val_accuracy: 0.8826 - val_dice_coef: 0.1517 - val_iou: 0.0822 - val_recall_1: 0.0000e+00 - val_precision_1: 0.0000e+00 - lr: 1.0000e-05\n","Epoch 7/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4518 - accuracy: 0.9190 - dice_coef: 0.5482 - iou: 0.3788 - recall_1: 0.9739 - precision_1: 0.5719\n","Epoch 7: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 300ms/step - loss: 0.4518 - accuracy: 0.9190 - dice_coef: 0.5482 - iou: 0.3788 - recall_1: 0.9739 - precision_1: 0.5719 - val_loss: 0.8550 - val_accuracy: 0.8826 - val_dice_coef: 0.1448 - val_iou: 0.0781 - val_recall_1: 0.0000e+00 - val_precision_1: 0.0000e+00 - lr: 1.0000e-06\n","Epoch 8/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4485 - accuracy: 0.9208 - dice_coef: 0.5515 - iou: 0.3824 - recall_1: 0.9752 - precision_1: 0.5774\n","Epoch 8: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 294ms/step - loss: 0.4485 - accuracy: 0.9208 - dice_coef: 0.5515 - iou: 0.3824 - recall_1: 0.9752 - precision_1: 0.5774 - val_loss: 0.8604 - val_accuracy: 0.8826 - val_dice_coef: 0.1385 - val_iou: 0.0744 - val_recall_1: 0.0000e+00 - val_precision_1: 0.0000e+00 - lr: 1.0000e-06\n","Epoch 9/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4433 - accuracy: 0.9236 - dice_coef: 0.5567 - iou: 0.3870 - recall_1: 0.9759 - precision_1: 0.5865\n","Epoch 9: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4433 - accuracy: 0.9236 - dice_coef: 0.5567 - iou: 0.3870 - recall_1: 0.9759 - precision_1: 0.5865 - val_loss: 0.8615 - val_accuracy: 0.8826 - val_dice_coef: 0.1372 - val_iou: 0.0737 - val_recall_1: 7.7997e-05 - val_precision_1: 0.8571 - lr: 1.0000e-06\n","Epoch 10/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4417 - accuracy: 0.9245 - dice_coef: 0.5583 - iou: 0.3889 - recall_1: 0.9743 - precision_1: 0.5896\n","Epoch 10: val_loss did not improve from 0.81148\n","30/30 [==============================] - 9s 286ms/step - loss: 0.4417 - accuracy: 0.9245 - dice_coef: 0.5583 - iou: 0.3889 - recall_1: 0.9743 - precision_1: 0.5896 - val_loss: 0.8525 - val_accuracy: 0.8830 - val_dice_coef: 0.1472 - val_iou: 0.0795 - val_recall_1: 0.0079 - val_precision_1: 0.6441 - lr: 1.0000e-06\n","Epoch 11/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4371 - accuracy: 0.9264 - dice_coef: 0.5629 - iou: 0.3930 - recall_1: 0.9753 - precision_1: 0.5958\n","Epoch 11: val_loss did not improve from 0.81148\n","\n","Epoch 11: ReduceLROnPlateau reducing learning rate to 1e-07.\n","30/30 [==============================] - 9s 286ms/step - loss: 0.4371 - accuracy: 0.9264 - dice_coef: 0.5629 - iou: 0.3930 - recall_1: 0.9753 - precision_1: 0.5958 - val_loss: 0.8220 - val_accuracy: 0.8849 - val_dice_coef: 0.1777 - val_iou: 0.0977 - val_recall_1: 0.0411 - val_precision_1: 0.6562 - lr: 1.0000e-06\n","Epoch 12/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4370 - accuracy: 0.9276 - dice_coef: 0.5630 - iou: 0.3938 - recall_1: 0.9747 - precision_1: 0.6002\n","Epoch 12: val_loss improved from 0.81148 to 0.76033, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 329ms/step - loss: 0.4370 - accuracy: 0.9276 - dice_coef: 0.5630 - iou: 0.3938 - recall_1: 0.9747 - precision_1: 0.6002 - val_loss: 0.7603 - val_accuracy: 0.8932 - val_dice_coef: 0.2332 - val_iou: 0.1336 - val_recall_1: 0.1531 - val_precision_1: 0.7078 - lr: 1.0000e-07\n","Epoch 13/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4357 - accuracy: 0.9275 - dice_coef: 0.5643 - iou: 0.3946 - recall_1: 0.9751 - precision_1: 0.5995\n","Epoch 13: val_loss improved from 0.76033 to 0.66153, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 331ms/step - loss: 0.4357 - accuracy: 0.9275 - dice_coef: 0.5643 - iou: 0.3946 - recall_1: 0.9751 - precision_1: 0.5995 - val_loss: 0.6615 - val_accuracy: 0.9100 - val_dice_coef: 0.3415 - val_iou: 0.2072 - val_recall_1: 0.3431 - val_precision_1: 0.7569 - lr: 1.0000e-07\n","Epoch 14/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4388 - accuracy: 0.9270 - dice_coef: 0.5612 - iou: 0.3925 - recall_1: 0.9732 - precision_1: 0.5980\n","Epoch 14: val_loss improved from 0.66153 to 0.56837, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 330ms/step - loss: 0.4388 - accuracy: 0.9270 - dice_coef: 0.5612 - iou: 0.3925 - recall_1: 0.9732 - precision_1: 0.5980 - val_loss: 0.5684 - val_accuracy: 0.9263 - val_dice_coef: 0.4279 - val_iou: 0.2739 - val_recall_1: 0.5445 - val_precision_1: 0.7593 - lr: 1.0000e-07\n","Epoch 15/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4350 - accuracy: 0.9285 - dice_coef: 0.5650 - iou: 0.3953 - recall_1: 0.9742 - precision_1: 0.6030\n","Epoch 15: val_loss improved from 0.56837 to 0.50861, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 332ms/step - loss: 0.4350 - accuracy: 0.9285 - dice_coef: 0.5650 - iou: 0.3953 - recall_1: 0.9742 - precision_1: 0.6030 - val_loss: 0.5086 - val_accuracy: 0.9333 - val_dice_coef: 0.4901 - val_iou: 0.3271 - val_recall_1: 0.6629 - val_precision_1: 0.7417 - lr: 1.0000e-07\n","Epoch 16/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4335 - accuracy: 0.9289 - dice_coef: 0.5665 - iou: 0.3963 - recall_1: 0.9742 - precision_1: 0.6046\n","Epoch 16: val_loss improved from 0.50861 to 0.46564, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 334ms/step - loss: 0.4335 - accuracy: 0.9289 - dice_coef: 0.5665 - iou: 0.3963 - recall_1: 0.9742 - precision_1: 0.6046 - val_loss: 0.4656 - val_accuracy: 0.9388 - val_dice_coef: 0.5416 - val_iou: 0.3741 - val_recall_1: 0.7581 - val_precision_1: 0.7308 - lr: 1.0000e-07\n","Epoch 17/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4349 - accuracy: 0.9289 - dice_coef: 0.5651 - iou: 0.3953 - recall_1: 0.9722 - precision_1: 0.6049\n","Epoch 17: val_loss improved from 0.46564 to 0.42987, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 338ms/step - loss: 0.4349 - accuracy: 0.9289 - dice_coef: 0.5651 - iou: 0.3953 - recall_1: 0.9722 - precision_1: 0.6049 - val_loss: 0.4299 - val_accuracy: 0.9439 - val_dice_coef: 0.5704 - val_iou: 0.3999 - val_recall_1: 0.8428 - val_precision_1: 0.7242 - lr: 1.0000e-07\n","Epoch 18/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4336 - accuracy: 0.9295 - dice_coef: 0.5664 - iou: 0.3973 - recall_1: 0.9750 - precision_1: 0.6066\n","Epoch 18: val_loss improved from 0.42987 to 0.40593, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 333ms/step - loss: 0.4336 - accuracy: 0.9295 - dice_coef: 0.5664 - iou: 0.3973 - recall_1: 0.9750 - precision_1: 0.6066 - val_loss: 0.4059 - val_accuracy: 0.9458 - val_dice_coef: 0.5945 - val_iou: 0.4231 - val_recall_1: 0.9016 - val_precision_1: 0.7126 - lr: 1.0000e-07\n","Epoch 19/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4323 - accuracy: 0.9295 - dice_coef: 0.5677 - iou: 0.3977 - recall_1: 0.9760 - precision_1: 0.6064\n","Epoch 19: val_loss improved from 0.40593 to 0.39803, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 334ms/step - loss: 0.4323 - accuracy: 0.9295 - dice_coef: 0.5677 - iou: 0.3977 - recall_1: 0.9760 - precision_1: 0.6064 - val_loss: 0.3980 - val_accuracy: 0.9455 - val_dice_coef: 0.6003 - val_iou: 0.4303 - val_recall_1: 0.9373 - val_precision_1: 0.7000 - lr: 1.0000e-07\n","Epoch 20/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4354 - accuracy: 0.9287 - dice_coef: 0.5646 - iou: 0.3951 - recall_1: 0.9710 - precision_1: 0.6041\n","Epoch 20: val_loss improved from 0.39803 to 0.39117, saving model to files/Unet_model_skull.h5\n","30/30 [==============================] - 10s 335ms/step - loss: 0.4354 - accuracy: 0.9287 - dice_coef: 0.5646 - iou: 0.3951 - recall_1: 0.9710 - precision_1: 0.6041 - val_loss: 0.3912 - val_accuracy: 0.9438 - val_dice_coef: 0.6062 - val_iou: 0.4354 - val_recall_1: 0.9531 - val_precision_1: 0.6882 - lr: 1.0000e-07\n","Epoch 21/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4330 - accuracy: 0.9293 - dice_coef: 0.5670 - iou: 0.3967 - recall_1: 0.9747 - precision_1: 0.6059\n","Epoch 21: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4330 - accuracy: 0.9293 - dice_coef: 0.5670 - iou: 0.3967 - recall_1: 0.9747 - precision_1: 0.6059 - val_loss: 0.3924 - val_accuracy: 0.9420 - val_dice_coef: 0.6120 - val_iou: 0.4425 - val_recall_1: 0.9597 - val_precision_1: 0.6788 - lr: 1.0000e-07\n","Epoch 22/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4330 - accuracy: 0.9290 - dice_coef: 0.5670 - iou: 0.3969 - recall_1: 0.9745 - precision_1: 0.6050\n","Epoch 22: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4330 - accuracy: 0.9290 - dice_coef: 0.5670 - iou: 0.3969 - recall_1: 0.9745 - precision_1: 0.6050 - val_loss: 0.3919 - val_accuracy: 0.9406 - val_dice_coef: 0.6088 - val_iou: 0.4386 - val_recall_1: 0.9625 - val_precision_1: 0.6726 - lr: 1.0000e-07\n","Epoch 23/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4345 - accuracy: 0.9286 - dice_coef: 0.5655 - iou: 0.3953 - recall_1: 0.9726 - precision_1: 0.6037\n","Epoch 23: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4345 - accuracy: 0.9286 - dice_coef: 0.5655 - iou: 0.3953 - recall_1: 0.9726 - precision_1: 0.6037 - val_loss: 0.3943 - val_accuracy: 0.9391 - val_dice_coef: 0.6060 - val_iou: 0.4362 - val_recall_1: 0.9639 - val_precision_1: 0.6662 - lr: 1.0000e-07\n","Epoch 24/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4316 - accuracy: 0.9309 - dice_coef: 0.5684 - iou: 0.3984 - recall_1: 0.9762 - precision_1: 0.6116\n","Epoch 24: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4316 - accuracy: 0.9309 - dice_coef: 0.5684 - iou: 0.3984 - recall_1: 0.9762 - precision_1: 0.6116 - val_loss: 0.3938 - val_accuracy: 0.9381 - val_dice_coef: 0.6031 - val_iou: 0.4322 - val_recall_1: 0.9646 - val_precision_1: 0.6621 - lr: 1.0000e-07\n","Epoch 25/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4293 - accuracy: 0.9313 - dice_coef: 0.5707 - iou: 0.4003 - recall_1: 0.9775 - precision_1: 0.6127\n","Epoch 25: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 292ms/step - loss: 0.4293 - accuracy: 0.9313 - dice_coef: 0.5707 - iou: 0.4003 - recall_1: 0.9775 - precision_1: 0.6127 - val_loss: 0.3973 - val_accuracy: 0.9370 - val_dice_coef: 0.6075 - val_iou: 0.4376 - val_recall_1: 0.9655 - val_precision_1: 0.6577 - lr: 1.0000e-07\n","Epoch 26/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4330 - accuracy: 0.9297 - dice_coef: 0.5670 - iou: 0.3969 - recall_1: 0.9731 - precision_1: 0.6077\n","Epoch 26: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 291ms/step - loss: 0.4330 - accuracy: 0.9297 - dice_coef: 0.5670 - iou: 0.3969 - recall_1: 0.9731 - precision_1: 0.6077 - val_loss: 0.3967 - val_accuracy: 0.9363 - val_dice_coef: 0.6075 - val_iou: 0.4369 - val_recall_1: 0.9659 - val_precision_1: 0.6551 - lr: 1.0000e-07\n","Epoch 27/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4309 - accuracy: 0.9307 - dice_coef: 0.5691 - iou: 0.3990 - recall_1: 0.9744 - precision_1: 0.6109\n","Epoch 27: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4309 - accuracy: 0.9307 - dice_coef: 0.5691 - iou: 0.3990 - recall_1: 0.9744 - precision_1: 0.6109 - val_loss: 0.3985 - val_accuracy: 0.9356 - val_dice_coef: 0.6021 - val_iou: 0.4322 - val_recall_1: 0.9663 - val_precision_1: 0.6524 - lr: 1.0000e-07\n","Epoch 28/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4308 - accuracy: 0.9306 - dice_coef: 0.5692 - iou: 0.3991 - recall_1: 0.9755 - precision_1: 0.6106\n","Epoch 28: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 291ms/step - loss: 0.4308 - accuracy: 0.9306 - dice_coef: 0.5692 - iou: 0.3991 - recall_1: 0.9755 - precision_1: 0.6106 - val_loss: 0.3985 - val_accuracy: 0.9354 - val_dice_coef: 0.6074 - val_iou: 0.4383 - val_recall_1: 0.9660 - val_precision_1: 0.6517 - lr: 1.0000e-07\n","Epoch 29/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4312 - accuracy: 0.9304 - dice_coef: 0.5688 - iou: 0.3994 - recall_1: 0.9754 - precision_1: 0.6099\n","Epoch 29: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4312 - accuracy: 0.9304 - dice_coef: 0.5688 - iou: 0.3994 - recall_1: 0.9754 - precision_1: 0.6099 - val_loss: 0.3989 - val_accuracy: 0.9353 - val_dice_coef: 0.6044 - val_iou: 0.4348 - val_recall_1: 0.9659 - val_precision_1: 0.6513 - lr: 1.0000e-07\n","Epoch 30/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4298 - accuracy: 0.9311 - dice_coef: 0.5702 - iou: 0.4004 - recall_1: 0.9752 - precision_1: 0.6121\n","Epoch 30: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 312ms/step - loss: 0.4298 - accuracy: 0.9311 - dice_coef: 0.5702 - iou: 0.4004 - recall_1: 0.9752 - precision_1: 0.6121 - val_loss: 0.3990 - val_accuracy: 0.9351 - val_dice_coef: 0.6057 - val_iou: 0.4360 - val_recall_1: 0.9659 - val_precision_1: 0.6506 - lr: 1.0000e-07\n","Epoch 31/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4306 - accuracy: 0.9303 - dice_coef: 0.5694 - iou: 0.3997 - recall_1: 0.9734 - precision_1: 0.6098\n","Epoch 31: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4306 - accuracy: 0.9303 - dice_coef: 0.5694 - iou: 0.3997 - recall_1: 0.9734 - precision_1: 0.6098 - val_loss: 0.3984 - val_accuracy: 0.9347 - val_dice_coef: 0.6023 - val_iou: 0.4321 - val_recall_1: 0.9663 - val_precision_1: 0.6492 - lr: 1.0000e-07\n","Epoch 32/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4301 - accuracy: 0.9307 - dice_coef: 0.5699 - iou: 0.3997 - recall_1: 0.9737 - precision_1: 0.6111\n","Epoch 32: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4301 - accuracy: 0.9307 - dice_coef: 0.5699 - iou: 0.3997 - recall_1: 0.9737 - precision_1: 0.6111 - val_loss: 0.4020 - val_accuracy: 0.9348 - val_dice_coef: 0.6020 - val_iou: 0.4332 - val_recall_1: 0.9661 - val_precision_1: 0.6493 - lr: 1.0000e-07\n","Epoch 33/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4298 - accuracy: 0.9318 - dice_coef: 0.5702 - iou: 0.4009 - recall_1: 0.9726 - precision_1: 0.6152\n","Epoch 33: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 310ms/step - loss: 0.4298 - accuracy: 0.9318 - dice_coef: 0.5702 - iou: 0.4009 - recall_1: 0.9726 - precision_1: 0.6152 - val_loss: 0.3990 - val_accuracy: 0.9349 - val_dice_coef: 0.5966 - val_iou: 0.4265 - val_recall_1: 0.9658 - val_precision_1: 0.6498 - lr: 1.0000e-07\n","Epoch 34/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4268 - accuracy: 0.9329 - dice_coef: 0.5732 - iou: 0.4030 - recall_1: 0.9761 - precision_1: 0.6185\n","Epoch 34: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4268 - accuracy: 0.9329 - dice_coef: 0.5732 - iou: 0.4030 - recall_1: 0.9761 - precision_1: 0.6185 - val_loss: 0.3978 - val_accuracy: 0.9349 - val_dice_coef: 0.6029 - val_iou: 0.4318 - val_recall_1: 0.9657 - val_precision_1: 0.6498 - lr: 1.0000e-07\n","Epoch 35/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4272 - accuracy: 0.9325 - dice_coef: 0.5728 - iou: 0.4024 - recall_1: 0.9752 - precision_1: 0.6174\n","Epoch 35: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4272 - accuracy: 0.9325 - dice_coef: 0.5728 - iou: 0.4024 - recall_1: 0.9752 - precision_1: 0.6174 - val_loss: 0.3975 - val_accuracy: 0.9349 - val_dice_coef: 0.5991 - val_iou: 0.4283 - val_recall_1: 0.9658 - val_precision_1: 0.6500 - lr: 1.0000e-07\n","Epoch 36/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4287 - accuracy: 0.9325 - dice_coef: 0.5713 - iou: 0.4012 - recall_1: 0.9742 - precision_1: 0.6173\n","Epoch 36: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4287 - accuracy: 0.9325 - dice_coef: 0.5713 - iou: 0.4012 - recall_1: 0.9742 - precision_1: 0.6173 - val_loss: 0.3992 - val_accuracy: 0.9352 - val_dice_coef: 0.6047 - val_iou: 0.4350 - val_recall_1: 0.9655 - val_precision_1: 0.6511 - lr: 1.0000e-07\n","Epoch 37/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4288 - accuracy: 0.9326 - dice_coef: 0.5712 - iou: 0.4014 - recall_1: 0.9736 - precision_1: 0.6181\n","Epoch 37: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4288 - accuracy: 0.9326 - dice_coef: 0.5712 - iou: 0.4014 - recall_1: 0.9736 - precision_1: 0.6181 - val_loss: 0.3995 - val_accuracy: 0.9352 - val_dice_coef: 0.5981 - val_iou: 0.4279 - val_recall_1: 0.9657 - val_precision_1: 0.6509 - lr: 1.0000e-07\n","Epoch 38/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4266 - accuracy: 0.9334 - dice_coef: 0.5734 - iou: 0.4034 - recall_1: 0.9742 - precision_1: 0.6206\n","Epoch 38: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4266 - accuracy: 0.9334 - dice_coef: 0.5734 - iou: 0.4034 - recall_1: 0.9742 - precision_1: 0.6206 - val_loss: 0.3983 - val_accuracy: 0.9353 - val_dice_coef: 0.6005 - val_iou: 0.4301 - val_recall_1: 0.9656 - val_precision_1: 0.6513 - lr: 1.0000e-07\n","Epoch 39/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4275 - accuracy: 0.9331 - dice_coef: 0.5725 - iou: 0.4025 - recall_1: 0.9746 - precision_1: 0.6195\n","Epoch 39: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4275 - accuracy: 0.9331 - dice_coef: 0.5725 - iou: 0.4025 - recall_1: 0.9746 - precision_1: 0.6195 - val_loss: 0.3972 - val_accuracy: 0.9352 - val_dice_coef: 0.6048 - val_iou: 0.4341 - val_recall_1: 0.9657 - val_precision_1: 0.6511 - lr: 1.0000e-07\n","Epoch 40/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4246 - accuracy: 0.9335 - dice_coef: 0.5754 - iou: 0.4051 - recall_1: 0.9764 - precision_1: 0.6208\n","Epoch 40: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 291ms/step - loss: 0.4246 - accuracy: 0.9335 - dice_coef: 0.5754 - iou: 0.4051 - recall_1: 0.9764 - precision_1: 0.6208 - val_loss: 0.3975 - val_accuracy: 0.9356 - val_dice_coef: 0.6019 - val_iou: 0.4310 - val_recall_1: 0.9653 - val_precision_1: 0.6527 - lr: 1.0000e-07\n","Epoch 41/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4272 - accuracy: 0.9319 - dice_coef: 0.5728 - iou: 0.4027 - recall_1: 0.9738 - precision_1: 0.6154\n","Epoch 41: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4272 - accuracy: 0.9319 - dice_coef: 0.5728 - iou: 0.4027 - recall_1: 0.9738 - precision_1: 0.6154 - val_loss: 0.3973 - val_accuracy: 0.9360 - val_dice_coef: 0.6005 - val_iou: 0.4298 - val_recall_1: 0.9647 - val_precision_1: 0.6542 - lr: 1.0000e-07\n","Epoch 42/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4255 - accuracy: 0.9327 - dice_coef: 0.5745 - iou: 0.4044 - recall_1: 0.9757 - precision_1: 0.6179\n","Epoch 42: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 291ms/step - loss: 0.4255 - accuracy: 0.9327 - dice_coef: 0.5745 - iou: 0.4044 - recall_1: 0.9757 - precision_1: 0.6179 - val_loss: 0.3987 - val_accuracy: 0.9361 - val_dice_coef: 0.6074 - val_iou: 0.4376 - val_recall_1: 0.9645 - val_precision_1: 0.6546 - lr: 1.0000e-07\n","Epoch 43/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4243 - accuracy: 0.9343 - dice_coef: 0.5757 - iou: 0.4057 - recall_1: 0.9748 - precision_1: 0.6239\n","Epoch 43: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 291ms/step - loss: 0.4243 - accuracy: 0.9343 - dice_coef: 0.5757 - iou: 0.4057 - recall_1: 0.9748 - precision_1: 0.6239 - val_loss: 0.3971 - val_accuracy: 0.9362 - val_dice_coef: 0.5992 - val_iou: 0.4291 - val_recall_1: 0.9644 - val_precision_1: 0.6551 - lr: 1.0000e-07\n","Epoch 44/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4283 - accuracy: 0.9323 - dice_coef: 0.5717 - iou: 0.4025 - recall_1: 0.9737 - precision_1: 0.6169\n","Epoch 44: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 290ms/step - loss: 0.4283 - accuracy: 0.9323 - dice_coef: 0.5717 - iou: 0.4025 - recall_1: 0.9737 - precision_1: 0.6169 - val_loss: 0.3985 - val_accuracy: 0.9365 - val_dice_coef: 0.6041 - val_iou: 0.4347 - val_recall_1: 0.9644 - val_precision_1: 0.6560 - lr: 1.0000e-07\n","Epoch 45/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4253 - accuracy: 0.9338 - dice_coef: 0.5747 - iou: 0.4046 - recall_1: 0.9738 - precision_1: 0.6224\n","Epoch 45: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4253 - accuracy: 0.9338 - dice_coef: 0.5747 - iou: 0.4046 - recall_1: 0.9738 - precision_1: 0.6224 - val_loss: 0.3980 - val_accuracy: 0.9363 - val_dice_coef: 0.6067 - val_iou: 0.4370 - val_recall_1: 0.9647 - val_precision_1: 0.6554 - lr: 1.0000e-07\n","Epoch 46/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4257 - accuracy: 0.9342 - dice_coef: 0.5743 - iou: 0.4051 - recall_1: 0.9745 - precision_1: 0.6237\n","Epoch 46: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 310ms/step - loss: 0.4257 - accuracy: 0.9342 - dice_coef: 0.5743 - iou: 0.4051 - recall_1: 0.9745 - precision_1: 0.6237 - val_loss: 0.3977 - val_accuracy: 0.9362 - val_dice_coef: 0.6015 - val_iou: 0.4313 - val_recall_1: 0.9652 - val_precision_1: 0.6549 - lr: 1.0000e-07\n","Epoch 47/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4248 - accuracy: 0.9347 - dice_coef: 0.5752 - iou: 0.4059 - recall_1: 0.9737 - precision_1: 0.6256\n","Epoch 47: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 289ms/step - loss: 0.4248 - accuracy: 0.9347 - dice_coef: 0.5752 - iou: 0.4059 - recall_1: 0.9737 - precision_1: 0.6256 - val_loss: 0.3959 - val_accuracy: 0.9364 - val_dice_coef: 0.6004 - val_iou: 0.4297 - val_recall_1: 0.9651 - val_precision_1: 0.6556 - lr: 1.0000e-07\n","Epoch 48/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4240 - accuracy: 0.9351 - dice_coef: 0.5760 - iou: 0.4065 - recall_1: 0.9742 - precision_1: 0.6270\n","Epoch 48: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 291ms/step - loss: 0.4240 - accuracy: 0.9351 - dice_coef: 0.5760 - iou: 0.4065 - recall_1: 0.9742 - precision_1: 0.6270 - val_loss: 0.3988 - val_accuracy: 0.9365 - val_dice_coef: 0.6042 - val_iou: 0.4350 - val_recall_1: 0.9651 - val_precision_1: 0.6560 - lr: 1.0000e-07\n","Epoch 49/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4237 - accuracy: 0.9345 - dice_coef: 0.5763 - iou: 0.4061 - recall_1: 0.9743 - precision_1: 0.6249\n","Epoch 49: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 294ms/step - loss: 0.4237 - accuracy: 0.9345 - dice_coef: 0.5763 - iou: 0.4061 - recall_1: 0.9743 - precision_1: 0.6249 - val_loss: 0.3973 - val_accuracy: 0.9370 - val_dice_coef: 0.6036 - val_iou: 0.4345 - val_recall_1: 0.9647 - val_precision_1: 0.6581 - lr: 1.0000e-07\n","Epoch 50/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4233 - accuracy: 0.9341 - dice_coef: 0.5767 - iou: 0.4068 - recall_1: 0.9733 - precision_1: 0.6235\n","Epoch 50: val_loss did not improve from 0.39117\n","30/30 [==============================] - 9s 292ms/step - loss: 0.4233 - accuracy: 0.9341 - dice_coef: 0.5767 - iou: 0.4068 - recall_1: 0.9733 - precision_1: 0.6235 - val_loss: 0.3940 - val_accuracy: 0.9373 - val_dice_coef: 0.6030 - val_iou: 0.4320 - val_recall_1: 0.9642 - val_precision_1: 0.6594 - lr: 1.0000e-07\n"]}]},{"cell_type":"markdown","source":["Output for The unet model"],"metadata":{"id":"ZJ0-0lw9mdgb"}},{"cell_type":"code","source":["import os\n","os.environ[\"TF_CPP_MIN_LOG_LEVEL\"] = \"2\"\n","import numpy as np\n","import cv2\n","from glob import glob\n","from tqdm import tqdm\n","import tensorflow as tf\n","from tensorflow.keras.utils import CustomObjectScope\n"," \n","if __name__ == \"__main__\":\n"," \n","    \"\"\" Loading model \"\"\"\n","    with CustomObjectScope({'iou': iou, 'dice_coef': dice_coef, 'dice_loss': dice_loss}):\n","        model = tf.keras.models.load_model(\"files/Unet_model_skull.h5\")\n","\n","    \"\"\" Dataset \"\"\"\n","    path = \"/content/drive/MyDrive/Colab Notebooks/Test skull\"\n","    test_img=sorted(glob(os.path.join(path,\"img\" ,\"*.png\")))\n","    ground_truth=sorted(glob(os.path.join(path,\"mask\" ,\"*.png\")))\n","\n","    \"\"\" Predicting the mask \"\"\"\n","    for x, y  in tqdm(zip(test_img,ground_truth), total=len(test_img)):\n","        \"\"\" Extracing the image name. \"\"\"\n","        image_name = x.split(\"/\")[-1]\n","\n","        \"\"\" Reading the image \"\"\"\n","        img = cv2.imread(x, cv2.IMREAD_COLOR)\n","        img = cv2.resize(img, (W, H))\n","        x = img/255.0\n","        x = x.astype(np.float32)\n","        x = np.expand_dims(x, axis=0)\n","\n","        \"\"\" Reading the mask \"\"\"\n","        img_y = cv2.imread(y, cv2.IMREAD_GRAYSCALE)\n","        img_y = cv2.resize(img_y, (W, H))\n","        img_y = np.expand_dims(img_y, axis=-1)  ## (512, 512, 1)\n","        img_y = np.concatenate([img_y, img_y, img_y], axis=-1)  ## (512, 512, 3)\n","\n","        \"\"\" Predicting the mask. \"\"\"\n","        y_pred = model.predict(x)[0] > 0.5\n","        y_pred = y_pred.astype(np.int32)\n","\n","        \"\"\" Saving the predicted mask along with the image and GT \"\"\"\n","        save_image_path = f\"/content/drive/MyDrive/Colab Notebooks/Results/{image_name}\"\n","        y_pred = np.concatenate([y_pred, y_pred, y_pred], axis=-1)\n","\n","        sep_line = np.ones((H, 10, 3)) * 255\n","\n","        output_image = np.concatenate([img, sep_line, img_y, sep_line, y_pred*255], axis=1)\n","        #cat_image = np.concatenate([sep_line,y_pred*255,sep_line], axis=1) # this line for just the output segment\n","        #cat_image = np.concatenate([ori_x,sep_line,y_pred*255], axis=1) # this line for just the output segment and the test image\n","        cv2.imwrite(save_image_path, output_image)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"k5KL4qR7XSx6","executionInfo":{"status":"ok","timestamp":1665129109503,"user_tz":-360,"elapsed":8632,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"e155e813-ecc7-4387-8643-4894c6e2342c"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["100%|██████████| 7/7 [00:05<00:00,  1.22it/s]\n"]}]},{"cell_type":"markdown","source":["Now lets load the pretrained weights"],"metadata":{"id":"st8AiQIgmjXo"}},{"cell_type":"code","source":[" \n","from tensorflow.keras.optimizers import Adam\n","import segmentation_models as sm\n","\n","input_shape = (256, 256, 3)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"BPPGP60JbbJ3","executionInfo":{"status":"ok","timestamp":1665129387630,"user_tz":-360,"elapsed":460,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"37da1670-8408-41a9-ff72-486215c8faf9"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Segmentation Models: using `keras` framework.\n"]}]},{"cell_type":"code","source":["pre_trained_unet_model = build_unet(input_shape)\n","pre_trained_unet_model.load_weights('unet_model_weights.h5')\n","pre_trained_unet_model_weights = pre_trained_unet_model.get_weights()[0][1]"],"metadata":{"id":"aEJCBwFpcVcB"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["pretrained_encoder_wts = np.load('pretrained_encoder-weights.npy')\n","\n","if pre_trained_unet_model_weights.all() == pretrained_encoder_wts.all():\n","    print(\"Both weights are identical\")\n","else: \n","    print(\"Something wrong, weghts are different\")"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"3HYLjqhSckxF","executionInfo":{"status":"ok","timestamp":1665130346189,"user_tz":-360,"elapsed":1343,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"b34062ce-245e-4fe3-e11d-1aa2200ca728"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Both weights are identical\n"]}]},{"cell_type":"markdown","source":["Now train the Unet model with the preprocessed weight"],"metadata":{"id":"9FNhGra5mpHJ"}},{"cell_type":"code","source":["if __name__ == \"__main__\":\n","    \"\"\" Seeding \"\"\"\n","    np.random.seed(42)\n","    tf.random.set_seed(42)\n","\n","    \"\"\" Directory for storing files \"\"\"\n","    create_dir(\"files\")\n","\n","    \"\"\" Hyperparameters \"\"\"\n","    batch_size =4\n","    lr = 1e-5\n","    num_epochs = 50\n","    model_path = os.path.join(\"files\", \"Unet_model_skull_pretrained.h5\")\n","    csv_path = os.path.join(\"files\", \"Skull_data_pretrained.csv\")\n","\n","    \"\"\" Dataset \"\"\"\n","    dataset_path = \"/content/drive/MyDrive/Colab Notebooks/Skull data\"\n","    (train_x, train_y), (valid_x, valid_y)  = load_data(dataset_path)\n","\n","    print(f\"Train: {len(train_x)} - {len(train_y)}\")\n","    print(f\"Valid: {len(valid_x)} - {len(valid_y)}\")\n","   \n","\n","    train_dataset = tf_dataset(train_x, train_y, batch=batch_size)\n","    valid_dataset = tf_dataset(valid_x, valid_y, batch=batch_size)\n","\n","    \"\"\" Model \"\"\"\n","    model = build_unet((H, W, 3))\n","    metrics = [dice_coef, iou, Recall(), Precision()]\n","    #model.compile(loss=dice_loss, optimizer=Adam(lr), metrics=metrics)\n","    # model.compile(loss=dice_loss, optimizer=Adam(lr), metrics=[\"accuracy\"])\n","    pre_trained_unet_model.compile(loss=dice_loss, optimizer=Adam(lr), metrics=[\"accuracy\",metrics])\n","    #pre_trained_unet_model.compile('Adam', loss=sm.losses.binary_focal_jaccard_loss, metrics=[sm.metrics.iou_score ,metrics])\n","\n","    callbacks = [\n","        ModelCheckpoint(model_path, verbose=1, save_best_only=True),\n","        ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, min_lr=1e-7, verbose=1),\n","        CSVLogger(csv_path)\n","    ]\n","\n","    pre_trained_unet_model.fit(\n","        train_dataset,\n","        epochs=num_epochs,\n","        validation_data=valid_dataset,\n","        callbacks=callbacks\n","    )"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"yANBWpz8glhq","executionInfo":{"status":"ok","timestamp":1665131187736,"user_tz":-360,"elapsed":523896,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"749f5180-c13e-4545-8c85-c1ee28173658"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Train: 120 - 120\n","Valid: 30 - 30\n","Epoch 1/50\n","30/30 [==============================] - ETA: 0s - loss: 0.7551 - accuracy: 0.5457 - dice_coef: 0.2449 - iou: 0.1400 - recall_4: 0.8726 - precision_4: 0.1752\n","Epoch 1: val_loss improved from inf to 0.79184, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 14s 351ms/step - loss: 0.7551 - accuracy: 0.5457 - dice_coef: 0.2449 - iou: 0.1400 - recall_4: 0.8726 - precision_4: 0.1752 - val_loss: 0.7918 - val_accuracy: 0.5954 - val_dice_coef: 0.2086 - val_iou: 0.1166 - val_recall_4: 0.8378 - val_precision_4: 0.2032 - lr: 1.0000e-05\n","Epoch 2/50\n","30/30 [==============================] - ETA: 0s - loss: 0.6801 - accuracy: 0.7436 - dice_coef: 0.3199 - iou: 0.1908 - recall_4: 0.9638 - precision_4: 0.2905\n","Epoch 2: val_loss improved from 0.79184 to 0.76687, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 334ms/step - loss: 0.6801 - accuracy: 0.7436 - dice_coef: 0.3199 - iou: 0.1908 - recall_4: 0.9638 - precision_4: 0.2905 - val_loss: 0.7669 - val_accuracy: 0.7063 - val_dice_coef: 0.2335 - val_iou: 0.1324 - val_recall_4: 0.9244 - val_precision_4: 0.2758 - lr: 1.0000e-05\n","Epoch 3/50\n","30/30 [==============================] - ETA: 0s - loss: 0.6290 - accuracy: 0.8216 - dice_coef: 0.3710 - iou: 0.2287 - recall_4: 0.9780 - precision_4: 0.3734\n","Epoch 3: val_loss improved from 0.76687 to 0.72149, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 351ms/step - loss: 0.6290 - accuracy: 0.8216 - dice_coef: 0.3710 - iou: 0.2287 - recall_4: 0.9780 - precision_4: 0.3734 - val_loss: 0.7215 - val_accuracy: 0.7940 - val_dice_coef: 0.2784 - val_iou: 0.1619 - val_recall_4: 0.9564 - val_precision_4: 0.3585 - lr: 1.0000e-05\n","Epoch 4/50\n","30/30 [==============================] - ETA: 0s - loss: 0.5863 - accuracy: 0.8607 - dice_coef: 0.4137 - iou: 0.2619 - recall_4: 0.9781 - precision_4: 0.4338\n","Epoch 4: val_loss improved from 0.72149 to 0.65604, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 349ms/step - loss: 0.5863 - accuracy: 0.8607 - dice_coef: 0.4137 - iou: 0.2619 - recall_4: 0.9781 - precision_4: 0.4338 - val_loss: 0.6560 - val_accuracy: 0.8453 - val_dice_coef: 0.3473 - val_iou: 0.2109 - val_recall_4: 0.9740 - val_precision_4: 0.4298 - lr: 1.0000e-05\n","Epoch 5/50\n","30/30 [==============================] - ETA: 0s - loss: 0.5452 - accuracy: 0.8896 - dice_coef: 0.4548 - iou: 0.2954 - recall_4: 0.9750 - precision_4: 0.4926\n","Epoch 5: val_loss improved from 0.65604 to 0.59657, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 341ms/step - loss: 0.5452 - accuracy: 0.8896 - dice_coef: 0.4548 - iou: 0.2954 - recall_4: 0.9750 - precision_4: 0.4926 - val_loss: 0.5966 - val_accuracy: 0.8922 - val_dice_coef: 0.3993 - val_iou: 0.2499 - val_recall_4: 0.9743 - val_precision_4: 0.5219 - lr: 1.0000e-05\n","Epoch 6/50\n","30/30 [==============================] - ETA: 0s - loss: 0.5051 - accuracy: 0.9140 - dice_coef: 0.4949 - iou: 0.3298 - recall_4: 0.9733 - precision_4: 0.5567\n","Epoch 6: val_loss improved from 0.59657 to 0.55399, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 335ms/step - loss: 0.5051 - accuracy: 0.9140 - dice_coef: 0.4949 - iou: 0.3298 - recall_4: 0.9733 - precision_4: 0.5567 - val_loss: 0.5540 - val_accuracy: 0.9041 - val_dice_coef: 0.4419 - val_iou: 0.2844 - val_recall_4: 0.9775 - val_precision_4: 0.5516 - lr: 1.0000e-05\n","Epoch 7/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4831 - accuracy: 0.9262 - dice_coef: 0.5169 - iou: 0.3499 - recall_4: 0.9493 - precision_4: 0.5984\n","Epoch 7: val_loss improved from 0.55399 to 0.53918, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 336ms/step - loss: 0.4831 - accuracy: 0.9262 - dice_coef: 0.5169 - iou: 0.3499 - recall_4: 0.9493 - precision_4: 0.5984 - val_loss: 0.5392 - val_accuracy: 0.8960 - val_dice_coef: 0.4591 - val_iou: 0.2987 - val_recall_4: 0.9777 - val_precision_4: 0.5309 - lr: 1.0000e-05\n","Epoch 8/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4584 - accuracy: 0.9352 - dice_coef: 0.5416 - iou: 0.3731 - recall_4: 0.9527 - precision_4: 0.6311\n","Epoch 8: val_loss improved from 0.53918 to 0.49986, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 336ms/step - loss: 0.4584 - accuracy: 0.9352 - dice_coef: 0.5416 - iou: 0.3731 - recall_4: 0.9527 - precision_4: 0.6311 - val_loss: 0.4999 - val_accuracy: 0.9178 - val_dice_coef: 0.4966 - val_iou: 0.3315 - val_recall_4: 0.9792 - val_precision_4: 0.5903 - lr: 1.0000e-05\n","Epoch 9/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4345 - accuracy: 0.9440 - dice_coef: 0.5655 - iou: 0.3954 - recall_4: 0.9536 - precision_4: 0.6672\n","Epoch 9: val_loss improved from 0.49986 to 0.46526, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 352ms/step - loss: 0.4345 - accuracy: 0.9440 - dice_coef: 0.5655 - iou: 0.3954 - recall_4: 0.9536 - precision_4: 0.6672 - val_loss: 0.4653 - val_accuracy: 0.9407 - val_dice_coef: 0.5304 - val_iou: 0.3621 - val_recall_4: 0.9599 - val_precision_4: 0.6734 - lr: 1.0000e-05\n","Epoch 10/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4200 - accuracy: 0.9483 - dice_coef: 0.5800 - iou: 0.4099 - recall_4: 0.9500 - precision_4: 0.6874\n","Epoch 10: val_loss improved from 0.46526 to 0.42933, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 331ms/step - loss: 0.4200 - accuracy: 0.9483 - dice_coef: 0.5800 - iou: 0.4099 - recall_4: 0.9500 - precision_4: 0.6874 - val_loss: 0.4293 - val_accuracy: 0.9455 - val_dice_coef: 0.5731 - val_iou: 0.4021 - val_recall_4: 0.9630 - val_precision_4: 0.6928 - lr: 1.0000e-05\n","Epoch 11/50\n","30/30 [==============================] - ETA: 0s - loss: 0.4049 - accuracy: 0.9532 - dice_coef: 0.5951 - iou: 0.4248 - recall_4: 0.9474 - precision_4: 0.7117\n","Epoch 11: val_loss improved from 0.42933 to 0.42062, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 335ms/step - loss: 0.4049 - accuracy: 0.9532 - dice_coef: 0.5951 - iou: 0.4248 - recall_4: 0.9474 - precision_4: 0.7117 - val_loss: 0.4206 - val_accuracy: 0.9398 - val_dice_coef: 0.5787 - val_iou: 0.4076 - val_recall_4: 0.9645 - val_precision_4: 0.6690 - lr: 1.0000e-05\n","Epoch 12/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3985 - accuracy: 0.9553 - dice_coef: 0.6015 - iou: 0.4318 - recall_4: 0.9454 - precision_4: 0.7230\n","Epoch 12: val_loss improved from 0.42062 to 0.39229, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 341ms/step - loss: 0.3985 - accuracy: 0.9553 - dice_coef: 0.6015 - iou: 0.4318 - recall_4: 0.9454 - precision_4: 0.7230 - val_loss: 0.3923 - val_accuracy: 0.9504 - val_dice_coef: 0.6081 - val_iou: 0.4379 - val_recall_4: 0.9586 - val_precision_4: 0.7154 - lr: 1.0000e-05\n","Epoch 13/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3913 - accuracy: 0.9569 - dice_coef: 0.6087 - iou: 0.4392 - recall_4: 0.9414 - precision_4: 0.7326\n","Epoch 13: val_loss improved from 0.39229 to 0.39065, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 339ms/step - loss: 0.3913 - accuracy: 0.9569 - dice_coef: 0.6087 - iou: 0.4392 - recall_4: 0.9414 - precision_4: 0.7326 - val_loss: 0.3906 - val_accuracy: 0.9442 - val_dice_coef: 0.6101 - val_iou: 0.4397 - val_recall_4: 0.9630 - val_precision_4: 0.6872 - lr: 1.0000e-05\n","Epoch 14/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3912 - accuracy: 0.9559 - dice_coef: 0.6088 - iou: 0.4403 - recall_4: 0.9391 - precision_4: 0.7284\n","Epoch 14: val_loss improved from 0.39065 to 0.38158, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 337ms/step - loss: 0.3912 - accuracy: 0.9559 - dice_coef: 0.6088 - iou: 0.4403 - recall_4: 0.9391 - precision_4: 0.7284 - val_loss: 0.3816 - val_accuracy: 0.9459 - val_dice_coef: 0.6198 - val_iou: 0.4498 - val_recall_4: 0.9647 - val_precision_4: 0.6937 - lr: 1.0000e-05\n","Epoch 15/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3783 - accuracy: 0.9592 - dice_coef: 0.6217 - iou: 0.4524 - recall_4: 0.9442 - precision_4: 0.7443\n","Epoch 15: val_loss improved from 0.38158 to 0.35821, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 332ms/step - loss: 0.3783 - accuracy: 0.9592 - dice_coef: 0.6217 - iou: 0.4524 - recall_4: 0.9442 - precision_4: 0.7443 - val_loss: 0.3582 - val_accuracy: 0.9538 - val_dice_coef: 0.6355 - val_iou: 0.4676 - val_recall_4: 0.9571 - val_precision_4: 0.7321 - lr: 1.0000e-05\n","Epoch 16/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3716 - accuracy: 0.9609 - dice_coef: 0.6284 - iou: 0.4593 - recall_4: 0.9480 - precision_4: 0.7520\n","Epoch 16: val_loss did not improve from 0.35821\n","30/30 [==============================] - 9s 286ms/step - loss: 0.3716 - accuracy: 0.9609 - dice_coef: 0.6284 - iou: 0.4593 - recall_4: 0.9480 - precision_4: 0.7520 - val_loss: 0.4017 - val_accuracy: 0.9264 - val_dice_coef: 0.6019 - val_iou: 0.4311 - val_recall_4: 0.9659 - val_precision_4: 0.6196 - lr: 1.0000e-05\n","Epoch 17/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3709 - accuracy: 0.9620 - dice_coef: 0.6291 - iou: 0.4605 - recall_4: 0.9409 - precision_4: 0.7609\n","Epoch 17: val_loss improved from 0.35821 to 0.35330, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 338ms/step - loss: 0.3709 - accuracy: 0.9620 - dice_coef: 0.6291 - iou: 0.4605 - recall_4: 0.9409 - precision_4: 0.7609 - val_loss: 0.3533 - val_accuracy: 0.9515 - val_dice_coef: 0.6451 - val_iou: 0.4766 - val_recall_4: 0.9596 - val_precision_4: 0.7201 - lr: 1.0000e-05\n","Epoch 18/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3654 - accuracy: 0.9625 - dice_coef: 0.6346 - iou: 0.4669 - recall_4: 0.9454 - precision_4: 0.7621\n","Epoch 18: val_loss improved from 0.35330 to 0.33430, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 329ms/step - loss: 0.3654 - accuracy: 0.9625 - dice_coef: 0.6346 - iou: 0.4669 - recall_4: 0.9454 - precision_4: 0.7621 - val_loss: 0.3343 - val_accuracy: 0.9597 - val_dice_coef: 0.6671 - val_iou: 0.5007 - val_recall_4: 0.9451 - val_precision_4: 0.7662 - lr: 1.0000e-05\n","Epoch 19/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3601 - accuracy: 0.9637 - dice_coef: 0.6399 - iou: 0.4717 - recall_4: 0.9457 - precision_4: 0.7692\n","Epoch 19: val_loss improved from 0.33430 to 0.32656, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 337ms/step - loss: 0.3601 - accuracy: 0.9637 - dice_coef: 0.6399 - iou: 0.4717 - recall_4: 0.9457 - precision_4: 0.7692 - val_loss: 0.3266 - val_accuracy: 0.9666 - val_dice_coef: 0.6733 - val_iou: 0.5084 - val_recall_4: 0.9180 - val_precision_4: 0.8195 - lr: 1.0000e-05\n","Epoch 20/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3634 - accuracy: 0.9628 - dice_coef: 0.6366 - iou: 0.4683 - recall_4: 0.9397 - precision_4: 0.7664\n","Epoch 20: val_loss did not improve from 0.32656\n","30/30 [==============================] - 9s 290ms/step - loss: 0.3634 - accuracy: 0.9628 - dice_coef: 0.6366 - iou: 0.4683 - recall_4: 0.9397 - precision_4: 0.7664 - val_loss: 0.3387 - val_accuracy: 0.9543 - val_dice_coef: 0.6593 - val_iou: 0.4921 - val_recall_4: 0.9540 - val_precision_4: 0.7355 - lr: 1.0000e-05\n","Epoch 21/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3563 - accuracy: 0.9643 - dice_coef: 0.6437 - iou: 0.4756 - recall_4: 0.9423 - precision_4: 0.7739\n","Epoch 21: val_loss did not improve from 0.32656\n","30/30 [==============================] - 9s 291ms/step - loss: 0.3563 - accuracy: 0.9643 - dice_coef: 0.6437 - iou: 0.4756 - recall_4: 0.9423 - precision_4: 0.7739 - val_loss: 0.3328 - val_accuracy: 0.9605 - val_dice_coef: 0.6687 - val_iou: 0.5030 - val_recall_4: 0.9397 - val_precision_4: 0.7727 - lr: 1.0000e-05\n","Epoch 22/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3504 - accuracy: 0.9659 - dice_coef: 0.6496 - iou: 0.4821 - recall_4: 0.9495 - precision_4: 0.7801\n","Epoch 22: val_loss did not improve from 0.32656\n","30/30 [==============================] - 9s 292ms/step - loss: 0.3504 - accuracy: 0.9659 - dice_coef: 0.6496 - iou: 0.4821 - recall_4: 0.9495 - precision_4: 0.7801 - val_loss: 0.3287 - val_accuracy: 0.9602 - val_dice_coef: 0.6715 - val_iou: 0.5060 - val_recall_4: 0.9450 - val_precision_4: 0.7688 - lr: 1.0000e-05\n","Epoch 23/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3509 - accuracy: 0.9660 - dice_coef: 0.6491 - iou: 0.4817 - recall_4: 0.9456 - precision_4: 0.7826\n","Epoch 23: val_loss improved from 0.32656 to 0.32478, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 333ms/step - loss: 0.3509 - accuracy: 0.9660 - dice_coef: 0.6491 - iou: 0.4817 - recall_4: 0.9456 - precision_4: 0.7826 - val_loss: 0.3248 - val_accuracy: 0.9632 - val_dice_coef: 0.6748 - val_iou: 0.5100 - val_recall_4: 0.9373 - val_precision_4: 0.7890 - lr: 1.0000e-05\n","Epoch 24/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3459 - accuracy: 0.9678 - dice_coef: 0.6541 - iou: 0.4871 - recall_4: 0.9491 - precision_4: 0.7917\n","Epoch 24: val_loss improved from 0.32478 to 0.31556, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 334ms/step - loss: 0.3459 - accuracy: 0.9678 - dice_coef: 0.6541 - iou: 0.4871 - recall_4: 0.9491 - precision_4: 0.7917 - val_loss: 0.3156 - val_accuracy: 0.9638 - val_dice_coef: 0.6830 - val_iou: 0.5188 - val_recall_4: 0.9385 - val_precision_4: 0.7920 - lr: 1.0000e-05\n","Epoch 25/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3428 - accuracy: 0.9676 - dice_coef: 0.6572 - iou: 0.4902 - recall_4: 0.9508 - precision_4: 0.7900\n","Epoch 25: val_loss did not improve from 0.31556\n","30/30 [==============================] - 9s 289ms/step - loss: 0.3428 - accuracy: 0.9676 - dice_coef: 0.6572 - iou: 0.4902 - recall_4: 0.9508 - precision_4: 0.7900 - val_loss: 0.3189 - val_accuracy: 0.9653 - val_dice_coef: 0.6820 - val_iou: 0.5180 - val_recall_4: 0.9277 - val_precision_4: 0.8059 - lr: 1.0000e-05\n","Epoch 26/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3422 - accuracy: 0.9681 - dice_coef: 0.6578 - iou: 0.4913 - recall_4: 0.9483 - precision_4: 0.7942\n","Epoch 26: val_loss did not improve from 0.31556\n","30/30 [==============================] - 9s 288ms/step - loss: 0.3422 - accuracy: 0.9681 - dice_coef: 0.6578 - iou: 0.4913 - recall_4: 0.9483 - precision_4: 0.7942 - val_loss: 0.3229 - val_accuracy: 0.9635 - val_dice_coef: 0.6779 - val_iou: 0.5128 - val_recall_4: 0.9299 - val_precision_4: 0.7944 - lr: 1.0000e-05\n","Epoch 27/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3392 - accuracy: 0.9682 - dice_coef: 0.6608 - iou: 0.4946 - recall_4: 0.9487 - precision_4: 0.7944\n","Epoch 27: val_loss did not improve from 0.31556\n","30/30 [==============================] - 9s 288ms/step - loss: 0.3392 - accuracy: 0.9682 - dice_coef: 0.6608 - iou: 0.4946 - recall_4: 0.9487 - precision_4: 0.7944 - val_loss: 0.3189 - val_accuracy: 0.9667 - val_dice_coef: 0.6813 - val_iou: 0.5172 - val_recall_4: 0.9211 - val_precision_4: 0.8181 - lr: 1.0000e-05\n","Epoch 28/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3380 - accuracy: 0.9685 - dice_coef: 0.6620 - iou: 0.4959 - recall_4: 0.9479 - precision_4: 0.7971\n","Epoch 28: val_loss did not improve from 0.31556\n","30/30 [==============================] - 9s 288ms/step - loss: 0.3380 - accuracy: 0.9685 - dice_coef: 0.6620 - iou: 0.4959 - recall_4: 0.9479 - precision_4: 0.7971 - val_loss: 0.3213 - val_accuracy: 0.9618 - val_dice_coef: 0.6825 - val_iou: 0.5190 - val_recall_4: 0.9409 - val_precision_4: 0.7793 - lr: 1.0000e-05\n","Epoch 29/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3366 - accuracy: 0.9690 - dice_coef: 0.6634 - iou: 0.4980 - recall_4: 0.9485 - precision_4: 0.7998\n","Epoch 29: val_loss did not improve from 0.31556\n","\n","Epoch 29: ReduceLROnPlateau reducing learning rate to 9.999999747378752e-07.\n","30/30 [==============================] - 9s 287ms/step - loss: 0.3366 - accuracy: 0.9690 - dice_coef: 0.6634 - iou: 0.4980 - recall_4: 0.9485 - precision_4: 0.7998 - val_loss: 0.3170 - val_accuracy: 0.9675 - val_dice_coef: 0.6863 - val_iou: 0.5233 - val_recall_4: 0.9141 - val_precision_4: 0.8272 - lr: 1.0000e-05\n","Epoch 30/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3321 - accuracy: 0.9704 - dice_coef: 0.6679 - iou: 0.5027 - recall_4: 0.9491 - precision_4: 0.8086\n","Epoch 30: val_loss improved from 0.31556 to 0.31441, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 336ms/step - loss: 0.3321 - accuracy: 0.9704 - dice_coef: 0.6679 - iou: 0.5027 - recall_4: 0.9491 - precision_4: 0.8086 - val_loss: 0.3144 - val_accuracy: 0.9670 - val_dice_coef: 0.6892 - val_iou: 0.5266 - val_recall_4: 0.9231 - val_precision_4: 0.8190 - lr: 1.0000e-06\n","Epoch 31/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3315 - accuracy: 0.9703 - dice_coef: 0.6685 - iou: 0.5033 - recall_4: 0.9513 - precision_4: 0.8064\n","Epoch 31: val_loss improved from 0.31441 to 0.31434, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 335ms/step - loss: 0.3315 - accuracy: 0.9703 - dice_coef: 0.6685 - iou: 0.5033 - recall_4: 0.9513 - precision_4: 0.8064 - val_loss: 0.3143 - val_accuracy: 0.9672 - val_dice_coef: 0.6859 - val_iou: 0.5223 - val_recall_4: 0.9230 - val_precision_4: 0.8200 - lr: 1.0000e-06\n","Epoch 32/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3307 - accuracy: 0.9705 - dice_coef: 0.6693 - iou: 0.5039 - recall_4: 0.9518 - precision_4: 0.8078\n","Epoch 32: val_loss did not improve from 0.31434\n","30/30 [==============================] - 9s 291ms/step - loss: 0.3307 - accuracy: 0.9705 - dice_coef: 0.6693 - iou: 0.5039 - recall_4: 0.9518 - precision_4: 0.8078 - val_loss: 0.3184 - val_accuracy: 0.9665 - val_dice_coef: 0.6844 - val_iou: 0.5215 - val_recall_4: 0.9269 - val_precision_4: 0.8138 - lr: 1.0000e-06\n","Epoch 33/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3321 - accuracy: 0.9706 - dice_coef: 0.6679 - iou: 0.5032 - recall_4: 0.9500 - precision_4: 0.8089\n","Epoch 33: val_loss did not improve from 0.31434\n","30/30 [==============================] - 9s 293ms/step - loss: 0.3321 - accuracy: 0.9706 - dice_coef: 0.6679 - iou: 0.5032 - recall_4: 0.9500 - precision_4: 0.8089 - val_loss: 0.3171 - val_accuracy: 0.9670 - val_dice_coef: 0.6823 - val_iou: 0.5184 - val_recall_4: 0.9231 - val_precision_4: 0.8188 - lr: 1.0000e-06\n","Epoch 34/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3281 - accuracy: 0.9713 - dice_coef: 0.6719 - iou: 0.5070 - recall_4: 0.9527 - precision_4: 0.8120\n","Epoch 34: val_loss did not improve from 0.31434\n","30/30 [==============================] - 9s 292ms/step - loss: 0.3281 - accuracy: 0.9713 - dice_coef: 0.6719 - iou: 0.5070 - recall_4: 0.9527 - precision_4: 0.8120 - val_loss: 0.3155 - val_accuracy: 0.9671 - val_dice_coef: 0.6852 - val_iou: 0.5214 - val_recall_4: 0.9236 - val_precision_4: 0.8189 - lr: 1.0000e-06\n","Epoch 35/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3296 - accuracy: 0.9704 - dice_coef: 0.6704 - iou: 0.5051 - recall_4: 0.9546 - precision_4: 0.8058\n","Epoch 35: val_loss improved from 0.31434 to 0.31425, saving model to files/Unet_model_skull_pretrained.h5\n","30/30 [==============================] - 10s 337ms/step - loss: 0.3296 - accuracy: 0.9704 - dice_coef: 0.6704 - iou: 0.5051 - recall_4: 0.9546 - precision_4: 0.8058 - val_loss: 0.3142 - val_accuracy: 0.9672 - val_dice_coef: 0.6861 - val_iou: 0.5224 - val_recall_4: 0.9237 - val_precision_4: 0.8196 - lr: 1.0000e-06\n","Epoch 36/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3311 - accuracy: 0.9708 - dice_coef: 0.6689 - iou: 0.5039 - recall_4: 0.9530 - precision_4: 0.8091\n","Epoch 36: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 291ms/step - loss: 0.3311 - accuracy: 0.9708 - dice_coef: 0.6689 - iou: 0.5039 - recall_4: 0.9530 - precision_4: 0.8091 - val_loss: 0.3167 - val_accuracy: 0.9672 - val_dice_coef: 0.6856 - val_iou: 0.5224 - val_recall_4: 0.9231 - val_precision_4: 0.8201 - lr: 1.0000e-06\n","Epoch 37/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3318 - accuracy: 0.9705 - dice_coef: 0.6682 - iou: 0.5033 - recall_4: 0.9513 - precision_4: 0.8077\n","Epoch 37: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 309ms/step - loss: 0.3318 - accuracy: 0.9705 - dice_coef: 0.6682 - iou: 0.5033 - recall_4: 0.9513 - precision_4: 0.8077 - val_loss: 0.3162 - val_accuracy: 0.9673 - val_dice_coef: 0.6816 - val_iou: 0.5176 - val_recall_4: 0.9239 - val_precision_4: 0.8200 - lr: 1.0000e-06\n","Epoch 38/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3300 - accuracy: 0.9709 - dice_coef: 0.6700 - iou: 0.5050 - recall_4: 0.9515 - precision_4: 0.8100\n","Epoch 38: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 289ms/step - loss: 0.3300 - accuracy: 0.9709 - dice_coef: 0.6700 - iou: 0.5050 - recall_4: 0.9515 - precision_4: 0.8100 - val_loss: 0.3153 - val_accuracy: 0.9680 - val_dice_coef: 0.6836 - val_iou: 0.5196 - val_recall_4: 0.9178 - val_precision_4: 0.8284 - lr: 1.0000e-06\n","Epoch 39/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3297 - accuracy: 0.9710 - dice_coef: 0.6704 - iou: 0.5055 - recall_4: 0.9540 - precision_4: 0.8096\n","Epoch 39: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 290ms/step - loss: 0.3297 - accuracy: 0.9710 - dice_coef: 0.6704 - iou: 0.5055 - recall_4: 0.9540 - precision_4: 0.8096 - val_loss: 0.3146 - val_accuracy: 0.9679 - val_dice_coef: 0.6858 - val_iou: 0.5221 - val_recall_4: 0.9185 - val_precision_4: 0.8270 - lr: 1.0000e-06\n","Epoch 40/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3274 - accuracy: 0.9715 - dice_coef: 0.6726 - iou: 0.5078 - recall_4: 0.9553 - precision_4: 0.8119\n","Epoch 40: val_loss did not improve from 0.31425\n","\n","Epoch 40: ReduceLROnPlateau reducing learning rate to 1e-07.\n","30/30 [==============================] - 9s 290ms/step - loss: 0.3274 - accuracy: 0.9715 - dice_coef: 0.6726 - iou: 0.5078 - recall_4: 0.9553 - precision_4: 0.8119 - val_loss: 0.3157 - val_accuracy: 0.9680 - val_dice_coef: 0.6823 - val_iou: 0.5182 - val_recall_4: 0.9170 - val_precision_4: 0.8285 - lr: 1.0000e-06\n","Epoch 41/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3268 - accuracy: 0.9715 - dice_coef: 0.6732 - iou: 0.5085 - recall_4: 0.9538 - precision_4: 0.8131\n","Epoch 41: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 291ms/step - loss: 0.3268 - accuracy: 0.9715 - dice_coef: 0.6732 - iou: 0.5085 - recall_4: 0.9538 - precision_4: 0.8131 - val_loss: 0.3159 - val_accuracy: 0.9678 - val_dice_coef: 0.6821 - val_iou: 0.5180 - val_recall_4: 0.9182 - val_precision_4: 0.8269 - lr: 1.0000e-07\n","Epoch 42/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3272 - accuracy: 0.9717 - dice_coef: 0.6728 - iou: 0.5081 - recall_4: 0.9546 - precision_4: 0.8137\n","Epoch 42: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 292ms/step - loss: 0.3272 - accuracy: 0.9717 - dice_coef: 0.6728 - iou: 0.5081 - recall_4: 0.9546 - precision_4: 0.8137 - val_loss: 0.3172 - val_accuracy: 0.9677 - val_dice_coef: 0.6869 - val_iou: 0.5238 - val_recall_4: 0.9189 - val_precision_4: 0.8258 - lr: 1.0000e-07\n","Epoch 43/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3284 - accuracy: 0.9713 - dice_coef: 0.6716 - iou: 0.5068 - recall_4: 0.9538 - precision_4: 0.8115\n","Epoch 43: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 291ms/step - loss: 0.3284 - accuracy: 0.9713 - dice_coef: 0.6716 - iou: 0.5068 - recall_4: 0.9538 - precision_4: 0.8115 - val_loss: 0.3170 - val_accuracy: 0.9677 - val_dice_coef: 0.6788 - val_iou: 0.5148 - val_recall_4: 0.9190 - val_precision_4: 0.8256 - lr: 1.0000e-07\n","Epoch 44/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3314 - accuracy: 0.9708 - dice_coef: 0.6686 - iou: 0.5045 - recall_4: 0.9519 - precision_4: 0.8093\n","Epoch 44: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 290ms/step - loss: 0.3314 - accuracy: 0.9708 - dice_coef: 0.6686 - iou: 0.5045 - recall_4: 0.9519 - precision_4: 0.8093 - val_loss: 0.3182 - val_accuracy: 0.9678 - val_dice_coef: 0.6836 - val_iou: 0.5199 - val_recall_4: 0.9184 - val_precision_4: 0.8265 - lr: 1.0000e-07\n","Epoch 45/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3289 - accuracy: 0.9713 - dice_coef: 0.6711 - iou: 0.5061 - recall_4: 0.9523 - precision_4: 0.8122\n","Epoch 45: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 291ms/step - loss: 0.3289 - accuracy: 0.9713 - dice_coef: 0.6711 - iou: 0.5061 - recall_4: 0.9523 - precision_4: 0.8122 - val_loss: 0.3173 - val_accuracy: 0.9677 - val_dice_coef: 0.6835 - val_iou: 0.5197 - val_recall_4: 0.9194 - val_precision_4: 0.8251 - lr: 1.0000e-07\n","Epoch 46/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3301 - accuracy: 0.9715 - dice_coef: 0.6699 - iou: 0.5056 - recall_4: 0.9537 - precision_4: 0.8131\n","Epoch 46: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 288ms/step - loss: 0.3301 - accuracy: 0.9715 - dice_coef: 0.6699 - iou: 0.5056 - recall_4: 0.9537 - precision_4: 0.8131 - val_loss: 0.3176 - val_accuracy: 0.9677 - val_dice_coef: 0.6847 - val_iou: 0.5211 - val_recall_4: 0.9193 - val_precision_4: 0.8254 - lr: 1.0000e-07\n","Epoch 47/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3306 - accuracy: 0.9710 - dice_coef: 0.6694 - iou: 0.5049 - recall_4: 0.9520 - precision_4: 0.8103\n","Epoch 47: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 290ms/step - loss: 0.3306 - accuracy: 0.9710 - dice_coef: 0.6694 - iou: 0.5049 - recall_4: 0.9520 - precision_4: 0.8103 - val_loss: 0.3163 - val_accuracy: 0.9678 - val_dice_coef: 0.6804 - val_iou: 0.5160 - val_recall_4: 0.9188 - val_precision_4: 0.8263 - lr: 1.0000e-07\n","Epoch 48/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3294 - accuracy: 0.9712 - dice_coef: 0.6706 - iou: 0.5061 - recall_4: 0.9532 - precision_4: 0.8112\n","Epoch 48: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 291ms/step - loss: 0.3294 - accuracy: 0.9712 - dice_coef: 0.6706 - iou: 0.5061 - recall_4: 0.9532 - precision_4: 0.8112 - val_loss: 0.3188 - val_accuracy: 0.9677 - val_dice_coef: 0.6827 - val_iou: 0.5191 - val_recall_4: 0.9198 - val_precision_4: 0.8248 - lr: 1.0000e-07\n","Epoch 49/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3288 - accuracy: 0.9714 - dice_coef: 0.6712 - iou: 0.5060 - recall_4: 0.9535 - precision_4: 0.8126\n","Epoch 49: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 292ms/step - loss: 0.3288 - accuracy: 0.9714 - dice_coef: 0.6712 - iou: 0.5060 - recall_4: 0.9535 - precision_4: 0.8126 - val_loss: 0.3187 - val_accuracy: 0.9677 - val_dice_coef: 0.6823 - val_iou: 0.5189 - val_recall_4: 0.9195 - val_precision_4: 0.8254 - lr: 1.0000e-07\n","Epoch 50/50\n","30/30 [==============================] - ETA: 0s - loss: 0.3278 - accuracy: 0.9711 - dice_coef: 0.6722 - iou: 0.5077 - recall_4: 0.9547 - precision_4: 0.8102\n","Epoch 50: val_loss did not improve from 0.31425\n","30/30 [==============================] - 9s 290ms/step - loss: 0.3278 - accuracy: 0.9711 - dice_coef: 0.6722 - iou: 0.5077 - recall_4: 0.9547 - precision_4: 0.8102 - val_loss: 0.3156 - val_accuracy: 0.9677 - val_dice_coef: 0.6822 - val_iou: 0.5179 - val_recall_4: 0.9195 - val_precision_4: 0.8253 - lr: 1.0000e-07\n"]}]},{"cell_type":"markdown","source":["Output for the trained model"],"metadata":{"id":"KiQhW-r-myKy"}},{"cell_type":"code","source":["import os\n","os.environ[\"TF_CPP_MIN_LOG_LEVEL\"] = \"2\"\n","import numpy as np\n","import cv2\n","from glob import glob\n","from tqdm import tqdm\n","import tensorflow as tf\n","from tensorflow.keras.utils import CustomObjectScope\n"," \n","if __name__ == \"__main__\":\n"," \n","    \"\"\" Loading model \"\"\"\n","    with CustomObjectScope({'iou': iou, 'dice_coef': dice_coef, 'dice_loss': dice_loss}):\n","        model = tf.keras.models.load_model(\"files/Unet_model_skull_pretrained.h5\")\n","\n","    \"\"\" Dataset \"\"\"\n","    path = \"/content/drive/MyDrive/Colab Notebooks/Test skull\"\n","    test_img=sorted(glob(os.path.join(path,\"img\" ,\"*.png\")))\n","    ground_truth=sorted(glob(os.path.join(path,\"mask\" ,\"*.png\")))\n","\n","    \"\"\" Predicting the mask \"\"\"\n","    for x, y  in tqdm(zip(test_img,ground_truth), total=len(test_img)):\n","        \"\"\" Extracing the image name. \"\"\"\n","        image_name = x.split(\"/\")[-1]\n","\n","        \"\"\" Reading the image \"\"\"\n","        img = cv2.imread(x, cv2.IMREAD_COLOR)\n","        img = cv2.resize(img, (W, H))\n","        x = img/255.0\n","        x = x.astype(np.float32)\n","        x = np.expand_dims(x, axis=0)\n","\n","        \"\"\" Reading the mask \"\"\"\n","        img_y = cv2.imread(y, cv2.IMREAD_GRAYSCALE)\n","        img_y = cv2.resize(img_y, (W, H))\n","        img_y = np.expand_dims(img_y, axis=-1)  ## (512, 512, 1)\n","        img_y = np.concatenate([img_y, img_y, img_y], axis=-1)  ## (512, 512, 3)\n","\n","        \"\"\" Predicting the mask. \"\"\"\n","        y_pred = model.predict(x)[0] > 0.5\n","        y_pred = y_pred.astype(np.int32)\n","\n","        \"\"\" Saving the predicted mask along with the image and GT \"\"\"\n","        save_image_path = f\"/content/drive/MyDrive/Colab Notebooks/Results/{image_name}\"\n","        y_pred = np.concatenate([y_pred, y_pred, y_pred], axis=-1)\n","\n","        sep_line = np.ones((H, 10, 3)) * 255\n","\n","        output_image = np.concatenate([img, sep_line, img_y, sep_line, y_pred*255], axis=1)\n","        #cat_image = np.concatenate([sep_line,y_pred*255,sep_line], axis=1) # this line for just the output segment\n","        #cat_image = np.concatenate([ori_x,sep_line,y_pred*255], axis=1) # this line for just the output segment and the test image\n","        cv2.imwrite(save_image_path, output_image)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"_eWdPjXKjZwR","executionInfo":{"status":"ok","timestamp":1665131617054,"user_tz":-360,"elapsed":3430,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"407a0f7c-b9f6-4b51-d08f-b5f8fbdf39e2"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["100%|██████████| 7/7 [00:01<00:00,  6.33it/s]\n"]}]},{"cell_type":"markdown","source":["Comparing the worst outputs"],"metadata":{"id":"tGVhSZ_boWzu"}},{"cell_type":"code","source":["x=cv2.imread(\"With.png\")\n","x=im_rgb = cv2.cvtColor(x, cv2.COLOR_BGR2RGB)\n","y=cv2.imread(\"Without.png\")\n","y=im_rgb = cv2.cvtColor(y, cv2.COLOR_BGR2RGB)\n","plt.subplot(1,2,1)\n","plt.imshow(x)\n","plt.title('pretrained')\n","plt.subplot(1,2,2)\n","plt.imshow(y)\n","plt.title('regular')\n","plt.show()\n"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":113},"id":"fNrLFQ7jlAj8","executionInfo":{"status":"ok","timestamp":1665132438246,"user_tz":-360,"elapsed":1282,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"dfeeedb2-5d39-4af6-dd41-e0b023169ada"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 432x288 with 2 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"markdown","source":["Comparing the IOU and EFFICIENCY graph"],"metadata":{"id":"jI6E5vs4otsb"}},{"cell_type":"code","source":["\n","import pandas as pd\n","import matplotlib.pyplot as plt\n","x=pd.read_csv(\"Skull_data.csv\")\n","y=pd.read_csv(\"Skull_data_pretrained.csv\")"],"metadata":{"id":"g0xCxjs4oFWy"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["regular_iou=x.iloc[:,3:4].values\n","pretrained_iou=y.iloc[:,3:4].values "],"metadata":{"id":"-I-ideqno1_o"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["epoch=x.iloc[:,0:1].values"],"metadata":{"id":"DTT2YL_UpFqC"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["plt.plot(epoch,regular_iou,label='regular iou')\n","plt.plot(epoch,pretrained_iou,label='pretrained iou')\n","plt.legend()\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":265},"id":"qHScNVXFpYpH","executionInfo":{"status":"ok","timestamp":1665132805823,"user_tz":-360,"elapsed":493,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"fcf6b681-227b-43ba-864c-566e52c8ff86"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 432x288 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"code","source":["regular_ef=x.iloc[:,1:2].values\n","pretrained_ef=y.iloc[:,1:2].values "],"metadata":{"id":"gEP4vgFUpjUB"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["plt.plot(epoch,regular_ef,label='regular ef')\n","plt.plot(epoch,pretrained_ef,label='pretrained ef')\n","plt.legend()\n","plt.show()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":265},"id":"FHXtATTAp06O","executionInfo":{"status":"ok","timestamp":1665132891849,"user_tz":-360,"elapsed":964,"user":{"displayName":"Shuvro Das","userId":"10178448084360053449"}},"outputId":"15c630cf-57d6-4a9a-e094-7ded0cf94f87"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 432x288 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}]},{"cell_type":"code","source":[],"metadata":{"id":"pNV7Hp_Np4Z-"},"execution_count":null,"outputs":[]}],"metadata":{"accelerator":"GPU","colab":{"collapsed_sections":[],"provenance":[],"mount_file_id":"1368vwNd4KxWtEBuqhLRMr43o0HEDvxz2","authorship_tag":"ABX9TyNhRnQz/croEoUnBLB3dI35"},"kernelspec":{"display_name":"Python 3","name":"python3"},"language_info":{"name":"python"}},"nbformat":4,"nbformat_minor":0}