Tamil_Character_Recognition.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Tamil Character Recognition.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8weaYD5QouNx",
        "outputId": "f5f7f718-4217-42c5-d331-1a491fb14b25"
      },
      "source": [
        "!pip install visualkeras"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting visualkeras\n",
            "  Downloading https://files.pythonhosted.org/packages/3c/7d/86ba86f710378e753ed690048710bf917b9121b7c6c5f5b07ec5323c4adc/visualkeras-0.0.1-py3-none-any.whl\n",
            "Requirement already satisfied: pillow>=6.2.0 in /usr/local/lib/python3.7/dist-packages (from visualkeras) (7.0.0)\n",
            "Installing collected packages: visualkeras\n",
            "Successfully installed visualkeras-0.0.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T6uJoZCMDNJE"
      },
      "source": [
        "from google.colab import files\n",
        "from zipfile import ZipFile\n",
        "import matplotlib.pyplot as plt\n",
        "import os\n",
        "import tensorflow as tf\n",
        "import keras\n",
        "import visualkeras\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "from keras.layers import Dense,Flatten,MaxPool2D,Conv2D,Dropout,GlobalAveragePooling2D,BatchNormalization,MaxPooling2D\n",
        "from keras.losses import binary_crossentropy,categorical_crossentropy\n",
        "from keras.models import Model,Sequential\n",
        "from keras.optimizers import RMSprop"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "resources": {
            "http://localhost:8080/nbextensions/google.colab/files.js": {
              "data": "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",
              "ok": true,
              "headers": [
                [
                  "content-type",
                  "application/javascript"
                ]
              ],
              "status": 200,
              "status_text": ""
            }
          },
          "base_uri": "https://localhost:8080/",
          "height": 72
        },
        "id": "cg41j5E6DYER",
        "outputId": "d7f5f933-3618-4940-bdcf-a1141441b5ac"
      },
      "source": [
        "files.upload()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "     <input type=\"file\" id=\"files-1c062658-2776-422c-a62a-5d03da144bdb\" name=\"files[]\" multiple disabled\n",
              "        style=\"border:none\" />\n",
              "     <output id=\"result-1c062658-2776-422c-a62a-5d03da144bdb\">\n",
              "      Upload widget is only available when the cell has been executed in the\n",
              "      current browser session. Please rerun this cell to enable.\n",
              "      </output>\n",
              "      <script src=\"/nbextensions/google.colab/files.js\"></script> "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Saving images.zip to images.zip\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qzeHe8j7DaiN"
      },
      "source": [
        "# Extract zipfile\n",
        "def extract_zip(fileName, targetFolder):\n",
        "    with ZipFile(fileName) as zipref:\n",
        "        zipref.extractall(targetFolder)\n",
        "\n",
        "extract_zip('/content/images.zip', '/content')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lI1eeFQ6ElZ0"
      },
      "source": [
        "BASE_DIR = '/content/images/'\n",
        "\n",
        "train_dir = BASE_DIR + 'train/'\n",
        "val_dir = BASE_DIR + 'val/'\n",
        "\n",
        "letters = ['a', 'aa', 'e', 'ee', 'u', 'uu', 'ea', 'eaa', 'i', 'o', 'oo']\n",
        "\n",
        "NUM_CLASSES = len(letters)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3Jlvm5H0lzwz"
      },
      "source": [
        "# from skimage.transform import resize\n",
        "# import matplotlib\n",
        "# import scipy.misc\n",
        "\n",
        "# image = plt.imread('/content/images/train/a/u16_000t04.tiff')\n",
        "# plt.imshow(image)\n",
        "# print(image.shape)\n",
        "\n",
        "# img_gray = rgb2gray(image[:,:,:3])\n",
        "# print(img_gray.shape)\n",
        "# image_res = resize(img_gray, (64,64))\n",
        "# print(image_res.shape)\n",
        "# plt.axis('off')\n",
        "# plt.imshow(image)\n",
        "# plt.savefig('test.png')\n",
        "# im = Image.fromarray(image_res)\n",
        "# im.save(\"aaaa.png\")\n",
        "\n",
        "\n",
        "# img_gray = rgb2gray(image[:,:,:3])\n",
        "# scipy.misc.toimage(img_gray).save('outfile.jpg')\n",
        "\n",
        "\n",
        "# matplotlib.image.imsave('name.png', image_res, cmap='gray')\n",
        "\n",
        "# print(img_gray.shape)\n",
        "# plt.imshow(img_gray, cmap='gray')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AZ1uZheczz7j",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "outputId": "a26bf0f9-70b7-4f08-81cc-bf9795f9ed39"
      },
      "source": [
        "img = plt.imread(train_dir + 'a/u16_000t04.jpeg')\n",
        "print(img.shape)\n",
        "plt.axis('off')\n",
        "plt.imshow(img)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(64, 64, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f0863c81cd0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K9eq31SXg5uX"
      },
      "source": [
        "train_datagen = ImageDataGenerator(\n",
        "    rotation_range=20, \n",
        "    width_shift_range=0.1, \n",
        "    height_shift_range=0.1, \n",
        "    shear_range=0.1, \n",
        "    zoom_range=0.1,  \n",
        "    rescale=1/255,\n",
        ")\n",
        "\n",
        "val_datagen = ImageDataGenerator(\n",
        "    rescale=1/255\n",
        ")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "znOi6AQYTXpW",
        "outputId": "37c37fc5-4a90-43fa-bbf3-c2cf57439627"
      },
      "source": [
        "train = train_datagen.flow_from_directory(train_dir, target_size=(64,64))\n",
        "validation = val_datagen.flow_from_directory(val_dir, target_size=(64,64))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 2420 images belonging to 11 classes.\n",
            "Found 579 images belonging to 11 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "eWurby9-cxDv",
        "outputId": "2caf8b4c-1b91-4ec0-8f36-199c73a7c9eb"
      },
      "source": [
        "img = plt.imread(train.directory + train.filenames[100])\n",
        "print(img.shape)\n",
        "plt.axis('off')\n",
        "plt.imshow(img)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(64, 64, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f0861c6e110>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "M1ch-65Hi4W2"
      },
      "source": [
        "ckp_filepath = 'trained-models/model'\n",
        "model_checkpoint = tf.keras.callbacks.ModelCheckpoint(filepath=ckp_filepath, \n",
        "                                                      monitor='val_mae', \n",
        "                                                      mode='auto',\n",
        "                                                      save_best_only=True, \n",
        "                                                      save_weights_only=True)\n",
        "\n",
        "reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(factor=0.9, monitor='val_mae', \n",
        "                                                 mode='auto', cooldown=0, \n",
        "                                                 patience=3, verbose=1, min_lr=1e-5)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xKmkii97hQg6"
      },
      "source": [
        "keras.backend.clear_session()\n",
        "model = Sequential([\n",
        "  Conv2D(16, 3, padding='same', activation='relu',input_shape=(64,64,3)),\n",
        "  MaxPooling2D(),\n",
        "  Conv2D(32, 3, padding='same', activation='relu'),\n",
        "  MaxPooling2D(),\n",
        "  Conv2D(64, 3, padding='same', activation='relu'),\n",
        "  MaxPooling2D(),\n",
        "  Dropout(0.2),\n",
        "  Conv2D(64, 3, padding='same', activation='relu'),\n",
        "  MaxPooling2D(),\n",
        "  Dropout(0.2),\n",
        "  Conv2D(64, 3, padding='same', activation='relu'),\n",
        "  MaxPooling2D(),\n",
        "  Dropout(0.2),\n",
        "  Conv2D(64, 3, padding='same', activation='relu'),\n",
        "  MaxPooling2D(),\n",
        "  Flatten(),\n",
        "  Dense(512, activation='relu'),\n",
        "  # Dense(128, activation='relu'),\n",
        "  # Dense(64, activation='relu'),\n",
        "  Dense(NUM_CLASSES, activation='softmax')\n",
        "])\n",
        "\n",
        "# visualkeras.layered_view(model, max_xy=700)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ifbHPANxPgG0"
      },
      "source": [
        "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['acc','mae'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "04Ek1epthOVf",
        "outputId": "56d1c5d4-6f75-48f1-ffdb-87643c5e406c"
      },
      "source": [
        "history = model.fit(train, epochs=30, validation_data=validation, callbacks=[model_checkpoint, reduce_lr])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "76/76 [==============================] - 5s 56ms/step - loss: 2.4025 - acc: 0.0686 - mae: 0.1653 - val_loss: 2.3978 - val_acc: 0.0950 - val_mae: 0.1653\n",
            "Epoch 2/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 2.3982 - acc: 0.0811 - mae: 0.1653 - val_loss: 2.3968 - val_acc: 0.0950 - val_mae: 0.1653\n",
            "Epoch 3/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 2.3559 - acc: 0.1182 - mae: 0.1637 - val_loss: 1.6769 - val_acc: 0.3437 - val_mae: 0.1290\n",
            "Epoch 4/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 1.6964 - acc: 0.3733 - mae: 0.1325 - val_loss: 1.1176 - val_acc: 0.5959 - val_mae: 0.0884\n",
            "Epoch 5/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 1.2565 - acc: 0.5566 - mae: 0.1057 - val_loss: 0.7290 - val_acc: 0.7461 - val_mae: 0.0610\n",
            "Epoch 6/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.9203 - acc: 0.6753 - mae: 0.0816 - val_loss: 0.4944 - val_acc: 0.8066 - val_mae: 0.0468\n",
            "Epoch 7/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.7109 - acc: 0.7576 - mae: 0.0659 - val_loss: 0.3731 - val_acc: 0.8584 - val_mae: 0.0352\n",
            "Epoch 8/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.6256 - acc: 0.7668 - mae: 0.0571 - val_loss: 0.3320 - val_acc: 0.8705 - val_mae: 0.0317\n",
            "Epoch 9/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.4856 - acc: 0.8256 - mae: 0.0454 - val_loss: 0.2742 - val_acc: 0.9085 - val_mae: 0.0259\n",
            "Epoch 10/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.4560 - acc: 0.8365 - mae: 0.0421 - val_loss: 0.1965 - val_acc: 0.9136 - val_mae: 0.0182\n",
            "Epoch 11/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.3753 - acc: 0.8637 - mae: 0.0349 - val_loss: 0.1835 - val_acc: 0.9309 - val_mae: 0.0191\n",
            "Epoch 12/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.3414 - acc: 0.8775 - mae: 0.0334 - val_loss: 0.1706 - val_acc: 0.9344 - val_mae: 0.0146\n",
            "Epoch 13/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.2849 - acc: 0.9029 - mae: 0.0269 - val_loss: 0.1670 - val_acc: 0.9447 - val_mae: 0.0134\n",
            "Epoch 14/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.2779 - acc: 0.8999 - mae: 0.0261 - val_loss: 0.1214 - val_acc: 0.9603 - val_mae: 0.0132\n",
            "Epoch 15/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.2298 - acc: 0.9135 - mae: 0.0234 - val_loss: 0.1213 - val_acc: 0.9585 - val_mae: 0.0110\n",
            "Epoch 16/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.2731 - acc: 0.9114 - mae: 0.0246 - val_loss: 0.0993 - val_acc: 0.9655 - val_mae: 0.0094\n",
            "Epoch 17/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.2311 - acc: 0.9193 - mae: 0.0219 - val_loss: 0.1227 - val_acc: 0.9482 - val_mae: 0.0114\n",
            "Epoch 18/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.2120 - acc: 0.9256 - mae: 0.0201 - val_loss: 0.0990 - val_acc: 0.9637 - val_mae: 0.0094\n",
            "Epoch 19/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.2433 - acc: 0.9169 - mae: 0.0220 - val_loss: 0.1863 - val_acc: 0.9378 - val_mae: 0.0136\n",
            "\n",
            "Epoch 00019: ReduceLROnPlateau reducing learning rate to 0.0009000000427477062.\n",
            "Epoch 20/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.1851 - acc: 0.9334 - mae: 0.0180 - val_loss: 0.1142 - val_acc: 0.9568 - val_mae: 0.0096\n",
            "Epoch 21/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.2013 - acc: 0.9286 - mae: 0.0187 - val_loss: 0.1031 - val_acc: 0.9637 - val_mae: 0.0084\n",
            "Epoch 22/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.1529 - acc: 0.9457 - mae: 0.0149 - val_loss: 0.0998 - val_acc: 0.9585 - val_mae: 0.0102\n",
            "Epoch 23/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.1627 - acc: 0.9450 - mae: 0.0156 - val_loss: 0.0653 - val_acc: 0.9827 - val_mae: 0.0061\n",
            "Epoch 24/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.1353 - acc: 0.9489 - mae: 0.0146 - val_loss: 0.0980 - val_acc: 0.9672 - val_mae: 0.0092\n",
            "Epoch 25/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.1358 - acc: 0.9559 - mae: 0.0133 - val_loss: 0.0889 - val_acc: 0.9775 - val_mae: 0.0076\n",
            "Epoch 26/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.1582 - acc: 0.9486 - mae: 0.0154 - val_loss: 0.0835 - val_acc: 0.9741 - val_mae: 0.0068\n",
            "\n",
            "Epoch 00026: ReduceLROnPlateau reducing learning rate to 0.0008100000384729356.\n",
            "Epoch 27/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.1372 - acc: 0.9484 - mae: 0.0136 - val_loss: 0.1055 - val_acc: 0.9706 - val_mae: 0.0084\n",
            "Epoch 28/30\n",
            "76/76 [==============================] - 4s 51ms/step - loss: 0.1075 - acc: 0.9622 - mae: 0.0115 - val_loss: 0.0764 - val_acc: 0.9775 - val_mae: 0.0062\n",
            "Epoch 29/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.1342 - acc: 0.9553 - mae: 0.0125 - val_loss: 0.1066 - val_acc: 0.9637 - val_mae: 0.0077\n",
            "\n",
            "Epoch 00029: ReduceLROnPlateau reducing learning rate to 0.0007290000503417104.\n",
            "Epoch 30/30\n",
            "76/76 [==============================] - 4s 50ms/step - loss: 0.1384 - acc: 0.9521 - mae: 0.0133 - val_loss: 0.0862 - val_acc: 0.9706 - val_mae: 0.0068\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r_enC_9MP__H",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 322
        },
        "outputId": "53effd97-9529-4e3b-d0f0-e8dd4299bb04"
      },
      "source": [
        "acc = history.history['acc']\n",
        "val_acc = history.history['val_acc']\n",
        "loss = history.history['loss']\n",
        "val_loss = history.history['val_loss']\n",
        "mae = history.history['mae']\n",
        "val_mae = history.history['val_mae']\n",
        "epochs = range(len(acc))\n",
        "\n",
        "fig,axes = plt.subplots(nrows=1,ncols=3, figsize=(20,5))\n",
        "\n",
        "axes[0].plot(epochs, acc, 'b', label='Training accuracy')\n",
        "axes[0].plot(epochs, val_acc, 'r', label='Validation accuracy')\n",
        "axes[0].legend(loc=0)\n",
        "\n",
        "axes[1].plot(epochs, loss, 'b', label='Training loss')\n",
        "axes[1].plot(epochs, val_loss, 'r', label='Validation loss')\n",
        "axes[1].legend(loc=0)\n",
        "\n",
        "axes[2].plot(epochs, mae, 'b', label='Training mae')\n",
        "axes[2].plot(epochs, val_mae, 'r', label='Validation mae')\n",
        "axes[2].legend(loc=0)\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1440x360 with 3 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gLDe3pwMg7YZ"
      },
      "source": [
        "# Prediction"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UpKmter4eMp9",
        "outputId": "26a48109-875b-458d-c343-98b58a16c7ab"
      },
      "source": [
        "prediction = model.predict_generator(train)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:1905: UserWarning: `Model.predict_generator` is deprecated and will be removed in a future version. Please use `Model.predict`, which supports generators.\n",
            "  warnings.warn('`Model.predict_generator` is deprecated and '\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "H8jhKfTK18G9",
        "outputId": "d6fe5436-0ccb-4d20-dc64-70daf5181d23"
      },
      "source": [
        "index = 245\n",
        "img = plt.imread(validation.directory + validation.filenames[index])\n",
        "plt.axis('off')\n",
        "plt.imshow(img)\n",
        "print(letters[prediction[index].argmax()])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "ea\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R1weKZe81t1l",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f2f4db26-6292-4523-fd54-9c72a5b7f209"
      },
      "source": [
        "model.evaluate(validation)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "19/19 [==============================] - 0s 11ms/step - loss: 0.0862 - acc: 0.9706 - mae: 0.0068\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.08624305576086044, 0.9706390500068665, 0.006792436353862286]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 60
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eDD_Sa99j5q2",
        "outputId": "c8bd1b30-c659-42ad-e15c-add66732b3a8"
      },
      "source": [
        "model.save('tamil_char_model')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: tamil_char_model/assets\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZwQmhUW0ztQA"
      },
      "source": [
        "# TFlite model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-WKZuk6Uzs1K",
        "outputId": "5d3e1372-3f76-42e7-f3a5-85e502768393"
      },
      "source": [
        "TAMIL_RECOG_MODEL = \"saved_models/tamil_recog\"\n",
        "tf.saved_model.save(model, TAMIL_RECOG_MODEL)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: saved_models/tamil_recog/assets\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tTqbA6DWlKes",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3ceded67-9be2-449f-c010-9a68b90179c8"
      },
      "source": [
        "import tensorflow_hub as hub\n",
        "tamil_model = hub.load(TAMIL_RECOG_MODEL)\n",
        "print(flowers_model)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<tensorflow.python.saved_model.load.Loader._recreate_base_user_object.<locals>._UserObject object at 0x7f0863ab31d0>\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ems1L-661c8P",
        "outputId": "1e63fd30-0dd5-4e98-c687-8d8eeff8dd1b"
      },
      "source": [
        "import numpy as np\n",
        "val_image_batch, val_label_batch = next(iter(validation))\n",
        "true_label_ids = np.argmax(val_label_batch, axis=-1)\n",
        "\n",
        "print(\"Validation batch shape:\", val_image_batch.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Validation batch shape: (32, 64, 64, 3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1ZcXemig1sSV",
        "outputId": "69a45639-02f0-4e2c-fd21-d86985ca8362"
      },
      "source": [
        "tf_model_predictions = tamil_model(val_image_batch)\n",
        "print(\"Prediction results shape:\", tf_model_predictions.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Prediction results shape: (32, 11)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QxkvF1aK2wEE"
      },
      "source": [
        "!mkdir \"tflite_models\""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RLPGiagD2B6X"
      },
      "source": [
        "TFLITE_MODEL = \"tflite_models/tamil_recog.tflite\"\n",
        "TFLITE_QUANT_MODEL = \"tflite_models/tamil_recog_quant.tflite\""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rQ8qbOaC2b4U",
        "outputId": "9ad47a24-8c80-4899-de48-d273bbffdb60"
      },
      "source": [
        "# Get the concrete function from the Keras model.\n",
        "run_model = tf.function(lambda x : flowers_model(x))\n",
        "\n",
        "# Save the concrete function.\n",
        "concrete_func = run_model.get_concrete_function(\n",
        "    tf.TensorSpec(model.inputs[0].shape, model.inputs[0].dtype)\n",
        ")\n",
        "\n",
        "# Convert the model\n",
        "converter = tf.lite.TFLiteConverter.from_concrete_functions([concrete_func])\n",
        "converted_tflite_model = converter.convert()\n",
        "open(TFLITE_MODEL, \"wb\").write(converted_tflite_model)\n",
        "\n",
        "# Convert the model to quantized version with post-training quantization\n",
        "converter = tf.lite.TFLiteConverter.from_concrete_functions([concrete_func])\n",
        "converter.optimizations = [tf.lite.Optimize.OPTIMIZE_FOR_SIZE]\n",
        "tflite_quant_model = converter.convert()\n",
        "open(TFLITE_QUANT_MODEL, \"wb\").write(tflite_quant_model)\n",
        "\n",
        "print(\"TFLite models and their sizes:\")\n",
        "!ls \"tflite_models\" -lh"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "TFLite models and their sizes:\n",
            "total 876K\n",
            "-rw-r--r-- 1 root root 188K Mar 30 06:38 tamil_recog_quant.tflite\n",
            "-rw-r--r-- 1 root root 686K Mar 30 06:38 tamil_recog.tflite\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}