VGG_transfer_Learning_.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "VGG transfer Learning .ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "metadata": {
        "id": "8c229G7v_8D8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1788
        },
        "outputId": "8f9b0ec7-d1b0-4457-ae84-5d1be0fa4733"
      },
      "cell_type": "code",
      "source": [
        "\n",
        "import numpy as np\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dropout, Flatten, Dense\n",
        "from keras import applications\n",
        "\n",
        "# dimensions of our images.\n",
        "img_width, img_height = 150, 150\n",
        "\n",
        "top_model_weights_path = 'bottleneck_fc_model.h5'\n",
        "train_data_dir = 'data/train'\n",
        "validation_data_dir = 'data/test'\n",
        "nb_train_samples = 2000\n",
        "nb_validation_samples = 400\n",
        "epochs = 50\n",
        "batch_size = 16\n",
        "\n",
        "\n",
        "def save_bottlebeck_features():\n",
        "    datagen = ImageDataGenerator(rescale=1. / 255)\n",
        "\n",
        "    # build the VGG16 network\n",
        "    model = applications.VGG16(include_top=False, weights='imagenet')\n",
        "\n",
        "    generator = datagen.flow_from_directory(\n",
        "        train_data_dir,\n",
        "        target_size=(img_width, img_height),\n",
        "        batch_size=batch_size,\n",
        "        class_mode=None,\n",
        "        shuffle=False)\n",
        "    bottleneck_features_train = model.predict_generator(\n",
        "        generator, nb_train_samples // batch_size)\n",
        "    np.save('bottleneck_features_train.npy',\n",
        "            bottleneck_features_train)\n",
        "\n",
        "    generator = datagen.flow_from_directory(\n",
        "        validation_data_dir,\n",
        "        target_size=(img_width, img_height),\n",
        "        batch_size=batch_size,\n",
        "        class_mode=None,\n",
        "        shuffle=False)\n",
        "    bottleneck_features_validation = model.predict_generator(\n",
        "        generator, nb_validation_samples // batch_size)\n",
        "    np.save('bottleneck_features_validation.npy',\n",
        "            bottleneck_features_validation)\n",
        "\n",
        "\n",
        "def train_top_model():\n",
        "    train_data = np.load('bottleneck_features_train.npy')\n",
        "    train_labels = np.array(\n",
        "        [0] * (nb_train_samples // 2) + [1] * (nb_train_samples // 2))\n",
        "\n",
        "    validation_data = np.load('bottleneck_features_validation.npy')\n",
        "    validation_labels = np.array(\n",
        "        [0] * (nb_validation_samples // 2) + [1] * (nb_validation_samples // 2))\n",
        "\n",
        "    model = Sequential()\n",
        "    model.add(Flatten(input_shape=train_data.shape[1:]))\n",
        "    model.add(Dense(256, activation='relu'))\n",
        "    model.add(Dropout(0.5))\n",
        "    model.add(Dense(1, activation='sigmoid'))\n",
        "\n",
        "    model.compile(optimizer='rmsprop',\n",
        "                  loss='binary_crossentropy', metrics=['accuracy'])\n",
        "\n",
        "    model.fit(train_data, train_labels,\n",
        "              epochs=epochs,\n",
        "              batch_size=batch_size,\n",
        "              validation_data=(validation_data, validation_labels))\n",
        "    model.save_weights(top_model_weights_path)\n",
        "\n",
        "\n",
        "save_bottlebeck_features()\n",
        "train_top_model()"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2000 images belonging to 2 classes.\n",
            "Found 400 images belonging to 2 classes.\n",
            "Train on 2000 samples, validate on 400 samples\n",
            "Epoch 1/50\n",
            "2000/2000 [==============================] - 5s 3ms/step - loss: 0.7536 - acc: 0.7685 - val_loss: 0.4975 - val_acc: 0.7525\n",
            "Epoch 2/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.3524 - acc: 0.8585 - val_loss: 0.2396 - val_acc: 0.8800\n",
            "Epoch 3/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.2900 - acc: 0.8800 - val_loss: 0.2205 - val_acc: 0.9025\n",
            "Epoch 4/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.2313 - acc: 0.9065 - val_loss: 0.3706 - val_acc: 0.8700\n",
            "Epoch 5/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.2241 - acc: 0.9140 - val_loss: 0.2489 - val_acc: 0.9075\n",
            "Epoch 6/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.2022 - acc: 0.9255 - val_loss: 0.2549 - val_acc: 0.9025\n",
            "Epoch 7/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.1745 - acc: 0.9365 - val_loss: 0.2643 - val_acc: 0.8925\n",
            "Epoch 8/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.1364 - acc: 0.9495 - val_loss: 0.3222 - val_acc: 0.8975\n",
            "Epoch 9/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.1441 - acc: 0.9390 - val_loss: 0.3692 - val_acc: 0.8725\n",
            "Epoch 10/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.1138 - acc: 0.9540 - val_loss: 0.4558 - val_acc: 0.8800\n",
            "Epoch 11/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.1052 - acc: 0.9590 - val_loss: 0.3433 - val_acc: 0.9000\n",
            "Epoch 12/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0989 - acc: 0.9640 - val_loss: 0.3529 - val_acc: 0.9100\n",
            "Epoch 13/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0944 - acc: 0.9655 - val_loss: 0.7508 - val_acc: 0.8425\n",
            "Epoch 14/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0633 - acc: 0.9740 - val_loss: 0.4111 - val_acc: 0.8975\n",
            "Epoch 15/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0757 - acc: 0.9740 - val_loss: 0.4224 - val_acc: 0.9125\n",
            "Epoch 16/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0664 - acc: 0.9760 - val_loss: 0.4873 - val_acc: 0.8925\n",
            "Epoch 17/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0620 - acc: 0.9755 - val_loss: 0.4661 - val_acc: 0.9100\n",
            "Epoch 18/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0503 - acc: 0.9795 - val_loss: 0.4732 - val_acc: 0.9050\n",
            "Epoch 19/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0440 - acc: 0.9825 - val_loss: 0.6062 - val_acc: 0.8850\n",
            "Epoch 20/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0535 - acc: 0.9815 - val_loss: 0.4799 - val_acc: 0.8975\n",
            "Epoch 21/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0317 - acc: 0.9885 - val_loss: 0.5343 - val_acc: 0.9100\n",
            "Epoch 22/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0463 - acc: 0.9825 - val_loss: 0.7639 - val_acc: 0.8825\n",
            "Epoch 23/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0338 - acc: 0.9875 - val_loss: 0.5676 - val_acc: 0.9000\n",
            "Epoch 24/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0372 - acc: 0.9870 - val_loss: 0.5921 - val_acc: 0.8975\n",
            "Epoch 25/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0273 - acc: 0.9915 - val_loss: 0.6080 - val_acc: 0.9000\n",
            "Epoch 26/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0324 - acc: 0.9895 - val_loss: 0.6025 - val_acc: 0.9075\n",
            "Epoch 27/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0195 - acc: 0.9935 - val_loss: 0.6188 - val_acc: 0.9050\n",
            "Epoch 28/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0249 - acc: 0.9920 - val_loss: 0.5630 - val_acc: 0.9025\n",
            "Epoch 29/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0256 - acc: 0.9920 - val_loss: 0.6885 - val_acc: 0.9025\n",
            "Epoch 30/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0175 - acc: 0.9955 - val_loss: 0.6631 - val_acc: 0.9100\n",
            "Epoch 31/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0219 - acc: 0.9920 - val_loss: 0.6925 - val_acc: 0.8900\n",
            "Epoch 32/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0266 - acc: 0.9925 - val_loss: 0.6972 - val_acc: 0.9050\n",
            "Epoch 33/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0258 - acc: 0.9925 - val_loss: 0.7132 - val_acc: 0.9000\n",
            "Epoch 34/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0222 - acc: 0.9915 - val_loss: 0.6989 - val_acc: 0.9025\n",
            "Epoch 35/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0161 - acc: 0.9935 - val_loss: 1.0063 - val_acc: 0.8725\n",
            "Epoch 36/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0101 - acc: 0.9965 - val_loss: 0.7829 - val_acc: 0.8950\n",
            "Epoch 37/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0159 - acc: 0.9930 - val_loss: 0.7440 - val_acc: 0.9050\n",
            "Epoch 38/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0293 - acc: 0.9920 - val_loss: 0.6652 - val_acc: 0.9050\n",
            "Epoch 39/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0044 - acc: 0.9985 - val_loss: 0.8297 - val_acc: 0.9100\n",
            "Epoch 40/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0118 - acc: 0.9960 - val_loss: 0.7843 - val_acc: 0.9125\n",
            "Epoch 41/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0102 - acc: 0.9975 - val_loss: 0.8226 - val_acc: 0.9025\n",
            "Epoch 42/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0375 - acc: 0.9910 - val_loss: 0.8019 - val_acc: 0.8975\n",
            "Epoch 43/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0148 - acc: 0.9960 - val_loss: 0.7739 - val_acc: 0.9075\n",
            "Epoch 44/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0145 - acc: 0.9955 - val_loss: 0.7667 - val_acc: 0.9025\n",
            "Epoch 45/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0103 - acc: 0.9965 - val_loss: 0.7644 - val_acc: 0.9000\n",
            "Epoch 46/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0144 - acc: 0.9965 - val_loss: 0.7403 - val_acc: 0.9100\n",
            "Epoch 47/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0082 - acc: 0.9980 - val_loss: 0.7318 - val_acc: 0.8975\n",
            "Epoch 48/50\n",
            "2000/2000 [==============================] - 4s 2ms/step - loss: 0.0164 - acc: 0.9960 - val_loss: 0.8060 - val_acc: 0.9100\n",
            "Epoch 49/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0173 - acc: 0.9955 - val_loss: 0.7735 - val_acc: 0.8950\n",
            "Epoch 50/50\n",
            "2000/2000 [==============================] - 5s 2ms/step - loss: 0.0093 - acc: 0.9965 - val_loss: 0.8408 - val_acc: 0.8925\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "NurL2GPdQWvp",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}