Sign_Langauge_Classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Sign_Langauge_Classification.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TxGoqEnaM0UG",
        "colab_type": "text"
      },
      "source": [
        "# Sign Language Image Classification\n",
        "\n",
        "In the below implemented convolution neural network, model classifies sign language digits for 10 classes from 0 to 9. I used MobileNet model, you can make the changes in the classes and model if required."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jH_gSxZwMeo-",
        "colab_type": "text"
      },
      "source": [
        "## Import the required libraries"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xGpdSV0wy7_w",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras import backend as k\n",
        "from tensorflow.keras.layers import Dense\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras.metrics import categorical_crossentropy\n",
        "from tensorflow.keras.preprocessing import image\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.applications import imagenet_utils\n",
        "from tensorflow.keras.applications.mobilenet import MobileNet\n",
        "from tensorflow.keras.applications.mobilenet import preprocess_input\n",
        "from sklearn.metrics import confusion_matrix\n",
        "import itertools\n",
        "import math\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline"
      ],
      "execution_count": 99,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zRM1anxgy9cc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# path for train, test and valid data of sign languages\n",
        "train_path = \"/content/gdrive/My Drive/datasets/train\"\n",
        "valid_path = \"/content/gdrive/My Drive/datasets/valid\"\n",
        "test_path = \"/content/gdrive/My Drive/datasets/test\""
      ],
      "execution_count": 100,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QJ_JcyXszNpk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        },
        "outputId": "c4de164a-e904-4071-9b4b-7e192c497824"
      },
      "source": [
        "# creating the batches for each path\n",
        "\n",
        "train_batches = ImageDataGenerator(rescale = 1.0/255.0, shear_range = 0.2, zoom_range = 0.2, horizontal_flip = True).flow_from_directory(train_path, target_size = (224, 224), batch_size = 10, class_mode = 'categorical')\n",
        "valid_batches = ImageDataGenerator(rescale = 1.0/255.0).flow_from_directory(valid_path, target_size = (224,224), batch_size = 10, class_mode = 'categorical')\n",
        "test_batches = ImageDataGenerator(rescale = 1.0/255.0).flow_from_directory(test_path, target_size = (224, 224), batch_size = 10, shuffle = False, class_mode = 'categorical')"
      ],
      "execution_count": 101,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found 1652 images belonging to 10 classes.\n",
            "Found 300 images belonging to 10 classes.\n",
            "Found 110 images belonging to 10 classes.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BWpepIPkMVom",
        "colab_type": "text"
      },
      "source": [
        "## Using the Mobile Net model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DzYGKDv20AhX",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "mobile = MobileNet()"
      ],
      "execution_count": 102,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "awYheCnV0E9T",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "b710479a-c292-424a-efa9-151c3e3d37ac"
      },
      "source": [
        "mobile.summary()"
      ],
      "execution_count": 103,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"mobilenet_1.00_224\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_6 (InputLayer)         [(None, 224, 224, 3)]     0         \n",
            "_________________________________________________________________\n",
            "conv1_pad (ZeroPadding2D)    (None, 225, 225, 3)       0         \n",
            "_________________________________________________________________\n",
            "conv1 (Conv2D)               (None, 112, 112, 32)      864       \n",
            "_________________________________________________________________\n",
            "conv1_bn (BatchNormalization (None, 112, 112, 32)      128       \n",
            "_________________________________________________________________\n",
            "conv1_relu (ReLU)            (None, 112, 112, 32)      0         \n",
            "_________________________________________________________________\n",
            "conv_dw_1 (DepthwiseConv2D)  (None, 112, 112, 32)      288       \n",
            "_________________________________________________________________\n",
            "conv_dw_1_bn (BatchNormaliza (None, 112, 112, 32)      128       \n",
            "_________________________________________________________________\n",
            "conv_dw_1_relu (ReLU)        (None, 112, 112, 32)      0         \n",
            "_________________________________________________________________\n",
            "conv_pw_1 (Conv2D)           (None, 112, 112, 64)      2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_1_bn (BatchNormaliza (None, 112, 112, 64)      256       \n",
            "_________________________________________________________________\n",
            "conv_pw_1_relu (ReLU)        (None, 112, 112, 64)      0         \n",
            "_________________________________________________________________\n",
            "conv_pad_2 (ZeroPadding2D)   (None, 113, 113, 64)      0         \n",
            "_________________________________________________________________\n",
            "conv_dw_2 (DepthwiseConv2D)  (None, 56, 56, 64)        576       \n",
            "_________________________________________________________________\n",
            "conv_dw_2_bn (BatchNormaliza (None, 56, 56, 64)        256       \n",
            "_________________________________________________________________\n",
            "conv_dw_2_relu (ReLU)        (None, 56, 56, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv_pw_2 (Conv2D)           (None, 56, 56, 128)       8192      \n",
            "_________________________________________________________________\n",
            "conv_pw_2_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_pw_2_relu (ReLU)        (None, 56, 56, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_3 (DepthwiseConv2D)  (None, 56, 56, 128)       1152      \n",
            "_________________________________________________________________\n",
            "conv_dw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_dw_3_relu (ReLU)        (None, 56, 56, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_3 (Conv2D)           (None, 56, 56, 128)       16384     \n",
            "_________________________________________________________________\n",
            "conv_pw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_pw_3_relu (ReLU)        (None, 56, 56, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_pad_4 (ZeroPadding2D)   (None, 57, 57, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_4 (DepthwiseConv2D)  (None, 28, 28, 128)       1152      \n",
            "_________________________________________________________________\n",
            "conv_dw_4_bn (BatchNormaliza (None, 28, 28, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_dw_4_relu (ReLU)        (None, 28, 28, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_4 (Conv2D)           (None, 28, 28, 256)       32768     \n",
            "_________________________________________________________________\n",
            "conv_pw_4_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_pw_4_relu (ReLU)        (None, 28, 28, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_5 (DepthwiseConv2D)  (None, 28, 28, 256)       2304      \n",
            "_________________________________________________________________\n",
            "conv_dw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_dw_5_relu (ReLU)        (None, 28, 28, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_5 (Conv2D)           (None, 28, 28, 256)       65536     \n",
            "_________________________________________________________________\n",
            "conv_pw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_pw_5_relu (ReLU)        (None, 28, 28, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_pad_6 (ZeroPadding2D)   (None, 29, 29, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_6 (DepthwiseConv2D)  (None, 14, 14, 256)       2304      \n",
            "_________________________________________________________________\n",
            "conv_dw_6_bn (BatchNormaliza (None, 14, 14, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_dw_6_relu (ReLU)        (None, 14, 14, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_6 (Conv2D)           (None, 14, 14, 512)       131072    \n",
            "_________________________________________________________________\n",
            "conv_pw_6_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_6_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_7 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_7_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_7 (Conv2D)           (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_7_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_8 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_8_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_8 (Conv2D)           (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_8_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_9 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_9_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_9 (Conv2D)           (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_9_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_10 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_10_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_10 (Conv2D)          (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_10_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_11 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_11_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_11 (Conv2D)          (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_11_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pad_12 (ZeroPadding2D)  (None, 15, 15, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_12 (DepthwiseConv2D) (None, 7, 7, 512)         4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_12_bn (BatchNormaliz (None, 7, 7, 512)         2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_12_relu (ReLU)       (None, 7, 7, 512)         0         \n",
            "_________________________________________________________________\n",
            "conv_pw_12 (Conv2D)          (None, 7, 7, 1024)        524288    \n",
            "_________________________________________________________________\n",
            "conv_pw_12_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
            "_________________________________________________________________\n",
            "conv_pw_12_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
            "_________________________________________________________________\n",
            "conv_dw_13 (DepthwiseConv2D) (None, 7, 7, 1024)        9216      \n",
            "_________________________________________________________________\n",
            "conv_dw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
            "_________________________________________________________________\n",
            "conv_dw_13_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
            "_________________________________________________________________\n",
            "conv_pw_13 (Conv2D)          (None, 7, 7, 1024)        1048576   \n",
            "_________________________________________________________________\n",
            "conv_pw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
            "_________________________________________________________________\n",
            "conv_pw_13_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
            "_________________________________________________________________\n",
            "global_average_pooling2d_5 ( (None, 1024)              0         \n",
            "_________________________________________________________________\n",
            "reshape_1 (Reshape)          (None, 1, 1, 1024)        0         \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 1, 1, 1024)        0         \n",
            "_________________________________________________________________\n",
            "conv_preds (Conv2D)          (None, 1, 1, 1000)        1025000   \n",
            "_________________________________________________________________\n",
            "reshape_2 (Reshape)          (None, 1000)              0         \n",
            "_________________________________________________________________\n",
            "predictions (Activation)     (None, 1000)              0         \n",
            "=================================================================\n",
            "Total params: 4,253,864\n",
            "Trainable params: 4,231,976\n",
            "Non-trainable params: 21,888\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Zc28Xqvx0IPu",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "x = mobile.layers[-6].output\n",
        "predictions = Dense(10, activation = \"softmax\")(x)\n",
        "model = Model(inputs = mobile.input, outputs = predictions)"
      ],
      "execution_count": 104,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZHkMCwH_0L6f",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "dfbb8058-c9ab-4a29-ed49-d1c15833b893"
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 105,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_5\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_6 (InputLayer)         [(None, 224, 224, 3)]     0         \n",
            "_________________________________________________________________\n",
            "conv1_pad (ZeroPadding2D)    (None, 225, 225, 3)       0         \n",
            "_________________________________________________________________\n",
            "conv1 (Conv2D)               (None, 112, 112, 32)      864       \n",
            "_________________________________________________________________\n",
            "conv1_bn (BatchNormalization (None, 112, 112, 32)      128       \n",
            "_________________________________________________________________\n",
            "conv1_relu (ReLU)            (None, 112, 112, 32)      0         \n",
            "_________________________________________________________________\n",
            "conv_dw_1 (DepthwiseConv2D)  (None, 112, 112, 32)      288       \n",
            "_________________________________________________________________\n",
            "conv_dw_1_bn (BatchNormaliza (None, 112, 112, 32)      128       \n",
            "_________________________________________________________________\n",
            "conv_dw_1_relu (ReLU)        (None, 112, 112, 32)      0         \n",
            "_________________________________________________________________\n",
            "conv_pw_1 (Conv2D)           (None, 112, 112, 64)      2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_1_bn (BatchNormaliza (None, 112, 112, 64)      256       \n",
            "_________________________________________________________________\n",
            "conv_pw_1_relu (ReLU)        (None, 112, 112, 64)      0         \n",
            "_________________________________________________________________\n",
            "conv_pad_2 (ZeroPadding2D)   (None, 113, 113, 64)      0         \n",
            "_________________________________________________________________\n",
            "conv_dw_2 (DepthwiseConv2D)  (None, 56, 56, 64)        576       \n",
            "_________________________________________________________________\n",
            "conv_dw_2_bn (BatchNormaliza (None, 56, 56, 64)        256       \n",
            "_________________________________________________________________\n",
            "conv_dw_2_relu (ReLU)        (None, 56, 56, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv_pw_2 (Conv2D)           (None, 56, 56, 128)       8192      \n",
            "_________________________________________________________________\n",
            "conv_pw_2_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_pw_2_relu (ReLU)        (None, 56, 56, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_3 (DepthwiseConv2D)  (None, 56, 56, 128)       1152      \n",
            "_________________________________________________________________\n",
            "conv_dw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_dw_3_relu (ReLU)        (None, 56, 56, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_3 (Conv2D)           (None, 56, 56, 128)       16384     \n",
            "_________________________________________________________________\n",
            "conv_pw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_pw_3_relu (ReLU)        (None, 56, 56, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_pad_4 (ZeroPadding2D)   (None, 57, 57, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_4 (DepthwiseConv2D)  (None, 28, 28, 128)       1152      \n",
            "_________________________________________________________________\n",
            "conv_dw_4_bn (BatchNormaliza (None, 28, 28, 128)       512       \n",
            "_________________________________________________________________\n",
            "conv_dw_4_relu (ReLU)        (None, 28, 28, 128)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_4 (Conv2D)           (None, 28, 28, 256)       32768     \n",
            "_________________________________________________________________\n",
            "conv_pw_4_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_pw_4_relu (ReLU)        (None, 28, 28, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_5 (DepthwiseConv2D)  (None, 28, 28, 256)       2304      \n",
            "_________________________________________________________________\n",
            "conv_dw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_dw_5_relu (ReLU)        (None, 28, 28, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_5 (Conv2D)           (None, 28, 28, 256)       65536     \n",
            "_________________________________________________________________\n",
            "conv_pw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_pw_5_relu (ReLU)        (None, 28, 28, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_pad_6 (ZeroPadding2D)   (None, 29, 29, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_6 (DepthwiseConv2D)  (None, 14, 14, 256)       2304      \n",
            "_________________________________________________________________\n",
            "conv_dw_6_bn (BatchNormaliza (None, 14, 14, 256)       1024      \n",
            "_________________________________________________________________\n",
            "conv_dw_6_relu (ReLU)        (None, 14, 14, 256)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_6 (Conv2D)           (None, 14, 14, 512)       131072    \n",
            "_________________________________________________________________\n",
            "conv_pw_6_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_6_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_7 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_7_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_7 (Conv2D)           (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_7_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_8 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_8_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_8 (Conv2D)           (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_8_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_9 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_9_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_9 (Conv2D)           (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_9_relu (ReLU)        (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_10 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_10_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_10 (Conv2D)          (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_10_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_11 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_11_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pw_11 (Conv2D)          (None, 14, 14, 512)       262144    \n",
            "_________________________________________________________________\n",
            "conv_pw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
            "_________________________________________________________________\n",
            "conv_pw_11_relu (ReLU)       (None, 14, 14, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_pad_12 (ZeroPadding2D)  (None, 15, 15, 512)       0         \n",
            "_________________________________________________________________\n",
            "conv_dw_12 (DepthwiseConv2D) (None, 7, 7, 512)         4608      \n",
            "_________________________________________________________________\n",
            "conv_dw_12_bn (BatchNormaliz (None, 7, 7, 512)         2048      \n",
            "_________________________________________________________________\n",
            "conv_dw_12_relu (ReLU)       (None, 7, 7, 512)         0         \n",
            "_________________________________________________________________\n",
            "conv_pw_12 (Conv2D)          (None, 7, 7, 1024)        524288    \n",
            "_________________________________________________________________\n",
            "conv_pw_12_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
            "_________________________________________________________________\n",
            "conv_pw_12_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
            "_________________________________________________________________\n",
            "conv_dw_13 (DepthwiseConv2D) (None, 7, 7, 1024)        9216      \n",
            "_________________________________________________________________\n",
            "conv_dw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
            "_________________________________________________________________\n",
            "conv_dw_13_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
            "_________________________________________________________________\n",
            "conv_pw_13 (Conv2D)          (None, 7, 7, 1024)        1048576   \n",
            "_________________________________________________________________\n",
            "conv_pw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
            "_________________________________________________________________\n",
            "conv_pw_13_relu (ReLU)       (None, 7, 7, 1024)        0         \n",
            "_________________________________________________________________\n",
            "global_average_pooling2d_5 ( (None, 1024)              0         \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 10)                10250     \n",
            "=================================================================\n",
            "Total params: 3,239,114\n",
            "Trainable params: 3,217,226\n",
            "Non-trainable params: 21,888\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t8TxNgPr0OIk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for layer in model.layers[:-24]:\n",
        "    layer.trainable = False"
      ],
      "execution_count": 106,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WDC-j0gEMPbC",
        "colab_type": "text"
      },
      "source": [
        "## Training the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jQA9xxa-0Ysp",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(optimizer = 'Adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])"
      ],
      "execution_count": 107,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YUhj1pw50d7-",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "fe8648ab-e2ca-46c8-a8e9-4b57a577aa33"
      },
      "source": [
        "model.fit_generator(train_batches, \n",
        "                    steps_per_epoch = math.ceil(train_batches.n / train_batches.batch_size),\n",
        "                    validation_data = valid_batches,\n",
        "                    validation_steps = math.ceil(valid_batches.n / valid_batches.batch_size),\n",
        "                    epochs = 50, verbose = 2 )"
      ],
      "execution_count": 108,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "166/166 - 23s - loss: 0.4490 - accuracy: 0.8590 - val_loss: 1.2257 - val_accuracy: 0.6967\n",
            "Epoch 2/50\n",
            "166/166 - 23s - loss: 0.1828 - accuracy: 0.9413 - val_loss: 0.7342 - val_accuracy: 0.8467\n",
            "Epoch 3/50\n",
            "166/166 - 23s - loss: 0.1045 - accuracy: 0.9685 - val_loss: 0.7378 - val_accuracy: 0.8333\n",
            "Epoch 4/50\n",
            "166/166 - 23s - loss: 0.0912 - accuracy: 0.9746 - val_loss: 0.2976 - val_accuracy: 0.9200\n",
            "Epoch 5/50\n",
            "166/166 - 23s - loss: 0.0472 - accuracy: 0.9849 - val_loss: 0.1451 - val_accuracy: 0.9800\n",
            "Epoch 6/50\n",
            "166/166 - 23s - loss: 0.0701 - accuracy: 0.9788 - val_loss: 0.2070 - val_accuracy: 0.9500\n",
            "Epoch 7/50\n",
            "166/166 - 23s - loss: 0.0809 - accuracy: 0.9782 - val_loss: 0.2029 - val_accuracy: 0.9600\n",
            "Epoch 8/50\n",
            "166/166 - 23s - loss: 0.0937 - accuracy: 0.9703 - val_loss: 0.2551 - val_accuracy: 0.9867\n",
            "Epoch 9/50\n",
            "166/166 - 23s - loss: 0.0252 - accuracy: 0.9933 - val_loss: 0.1875 - val_accuracy: 0.9900\n",
            "Epoch 10/50\n",
            "166/166 - 23s - loss: 0.0150 - accuracy: 0.9952 - val_loss: 0.1722 - val_accuracy: 0.9833\n",
            "Epoch 11/50\n",
            "166/166 - 23s - loss: 0.0382 - accuracy: 0.9867 - val_loss: 0.1820 - val_accuracy: 0.9800\n",
            "Epoch 12/50\n",
            "166/166 - 23s - loss: 0.0466 - accuracy: 0.9867 - val_loss: 2.2297 - val_accuracy: 0.6533\n",
            "Epoch 13/50\n",
            "166/166 - 23s - loss: 0.0393 - accuracy: 0.9915 - val_loss: 0.1486 - val_accuracy: 0.9767\n",
            "Epoch 14/50\n",
            "166/166 - 23s - loss: 0.0382 - accuracy: 0.9891 - val_loss: 0.4449 - val_accuracy: 0.9733\n",
            "Epoch 15/50\n",
            "166/166 - 23s - loss: 0.0119 - accuracy: 0.9952 - val_loss: 0.2527 - val_accuracy: 0.9867\n",
            "Epoch 16/50\n",
            "166/166 - 23s - loss: 0.0185 - accuracy: 0.9939 - val_loss: 0.2036 - val_accuracy: 0.9800\n",
            "Epoch 17/50\n",
            "166/166 - 23s - loss: 0.0325 - accuracy: 0.9873 - val_loss: 0.1034 - val_accuracy: 0.9767\n",
            "Epoch 18/50\n",
            "166/166 - 23s - loss: 0.0196 - accuracy: 0.9921 - val_loss: 0.0813 - val_accuracy: 0.9867\n",
            "Epoch 19/50\n",
            "166/166 - 23s - loss: 0.0199 - accuracy: 0.9946 - val_loss: 0.1027 - val_accuracy: 0.9867\n",
            "Epoch 20/50\n",
            "166/166 - 23s - loss: 0.0269 - accuracy: 0.9946 - val_loss: 0.0663 - val_accuracy: 0.9733\n",
            "Epoch 21/50\n",
            "166/166 - 23s - loss: 0.0454 - accuracy: 0.9855 - val_loss: 0.0703 - val_accuracy: 0.9800\n",
            "Epoch 22/50\n",
            "166/166 - 23s - loss: 0.0178 - accuracy: 0.9964 - val_loss: 0.0709 - val_accuracy: 0.9833\n",
            "Epoch 23/50\n",
            "166/166 - 23s - loss: 0.0026 - accuracy: 0.9988 - val_loss: 0.1476 - val_accuracy: 0.9767\n",
            "Epoch 24/50\n",
            "166/166 - 23s - loss: 0.0926 - accuracy: 0.9776 - val_loss: 0.2277 - val_accuracy: 0.9333\n",
            "Epoch 25/50\n",
            "166/166 - 23s - loss: 0.0276 - accuracy: 0.9927 - val_loss: 0.1048 - val_accuracy: 0.9833\n",
            "Epoch 26/50\n",
            "166/166 - 23s - loss: 0.0428 - accuracy: 0.9873 - val_loss: 0.1312 - val_accuracy: 0.9900\n",
            "Epoch 27/50\n",
            "166/166 - 23s - loss: 0.0387 - accuracy: 0.9879 - val_loss: 0.6798 - val_accuracy: 0.8367\n",
            "Epoch 28/50\n",
            "166/166 - 23s - loss: 0.0244 - accuracy: 0.9915 - val_loss: 0.2466 - val_accuracy: 0.9633\n",
            "Epoch 29/50\n",
            "166/166 - 23s - loss: 0.0244 - accuracy: 0.9909 - val_loss: 0.1967 - val_accuracy: 0.9800\n",
            "Epoch 30/50\n",
            "166/166 - 23s - loss: 0.0124 - accuracy: 0.9952 - val_loss: 0.2215 - val_accuracy: 0.9867\n",
            "Epoch 31/50\n",
            "166/166 - 23s - loss: 0.0152 - accuracy: 0.9952 - val_loss: 0.2225 - val_accuracy: 0.9867\n",
            "Epoch 32/50\n",
            "166/166 - 23s - loss: 4.6876e-04 - accuracy: 1.0000 - val_loss: 0.2399 - val_accuracy: 0.9900\n",
            "Epoch 33/50\n",
            "166/166 - 23s - loss: 0.0060 - accuracy: 0.9976 - val_loss: 0.3054 - val_accuracy: 0.9833\n",
            "Epoch 34/50\n",
            "166/166 - 24s - loss: 0.0279 - accuracy: 0.9927 - val_loss: 4.9045 - val_accuracy: 0.5533\n",
            "Epoch 35/50\n",
            "166/166 - 23s - loss: 0.0201 - accuracy: 0.9933 - val_loss: 0.2943 - val_accuracy: 0.9767\n",
            "Epoch 36/50\n",
            "166/166 - 23s - loss: 0.0105 - accuracy: 0.9952 - val_loss: 0.2381 - val_accuracy: 0.9900\n",
            "Epoch 37/50\n",
            "166/166 - 23s - loss: 0.0249 - accuracy: 0.9933 - val_loss: 0.4786 - val_accuracy: 0.9567\n",
            "Epoch 38/50\n",
            "166/166 - 23s - loss: 0.0204 - accuracy: 0.9933 - val_loss: 0.1566 - val_accuracy: 0.9900\n",
            "Epoch 39/50\n",
            "166/166 - 23s - loss: 0.0016 - accuracy: 0.9994 - val_loss: 0.1605 - val_accuracy: 0.9900\n",
            "Epoch 40/50\n",
            "166/166 - 23s - loss: 0.0171 - accuracy: 0.9958 - val_loss: 0.2604 - val_accuracy: 0.9833\n",
            "Epoch 41/50\n",
            "166/166 - 23s - loss: 0.0077 - accuracy: 0.9976 - val_loss: 0.1935 - val_accuracy: 0.9600\n",
            "Epoch 42/50\n",
            "166/166 - 23s - loss: 0.0602 - accuracy: 0.9837 - val_loss: 0.3795 - val_accuracy: 0.9600\n",
            "Epoch 43/50\n",
            "166/166 - 23s - loss: 0.0196 - accuracy: 0.9939 - val_loss: 0.1929 - val_accuracy: 0.9867\n",
            "Epoch 44/50\n",
            "166/166 - 23s - loss: 0.0509 - accuracy: 0.9873 - val_loss: 0.3137 - val_accuracy: 0.9733\n",
            "Epoch 45/50\n",
            "166/166 - 23s - loss: 0.0153 - accuracy: 0.9964 - val_loss: 0.1469 - val_accuracy: 0.9867\n",
            "Epoch 46/50\n",
            "166/166 - 23s - loss: 0.0114 - accuracy: 0.9964 - val_loss: 0.1030 - val_accuracy: 0.9900\n",
            "Epoch 47/50\n",
            "166/166 - 23s - loss: 0.0021 - accuracy: 0.9988 - val_loss: 0.1135 - val_accuracy: 0.9900\n",
            "Epoch 48/50\n",
            "166/166 - 23s - loss: 0.0184 - accuracy: 0.9952 - val_loss: 0.0925 - val_accuracy: 0.9900\n",
            "Epoch 49/50\n",
            "166/166 - 23s - loss: 0.0173 - accuracy: 0.9952 - val_loss: 0.1228 - val_accuracy: 0.9900\n",
            "Epoch 50/50\n",
            "166/166 - 23s - loss: 0.0118 - accuracy: 0.9964 - val_loss: 0.1804 - val_accuracy: 0.9900\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f6aa7b15080>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 108
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eiGRIXrBMHp3",
        "colab_type": "text"
      },
      "source": [
        "## Predictions using test data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "noQIDXFx0o0k",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test_labels = test_batches.classes"
      ],
      "execution_count": 109,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7hojOqiq0sMg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        },
        "outputId": "ba1955b1-352d-49a6-dff9-13cca0bf722e"
      },
      "source": [
        "test_labels"
      ],
      "execution_count": 110,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
              "       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,\n",
              "       4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,\n",
              "       6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,\n",
              "       8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9],\n",
              "      dtype=int32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 110
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hASnYkU20voQ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        },
        "outputId": "40e65e64-e81f-4ab3-b121-ac279442c01c"
      },
      "source": [
        "test_batches.class_indices"
      ],
      "execution_count": 111,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'0': 0,\n",
              " '1': 1,\n",
              " '2': 2,\n",
              " '3': 3,\n",
              " '4': 4,\n",
              " '5': 5,\n",
              " '6': 6,\n",
              " '7': 7,\n",
              " '8': 8,\n",
              " '9': 9}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 111
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NtPRTj5W02mb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "predictions = model.predict_generator(test_batches, steps = 11, verbose = 0)"
      ],
      "execution_count": 112,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G9vbrOYVL_DR",
        "colab_type": "text"
      },
      "source": [
        "## Plotting the test data by using confusion matrix"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GguFR4PS05Qy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# for confusion_matrix\n",
        "\n",
        "def plot_confusion_matrix(cm, classes,\n",
        "                         normalize = False,\n",
        "                         title = 'confusion_matrix',\n",
        "                         cmap = plt.cm.Blues):\n",
        "    plt.imshow(cm, interpolation = 'nearest', cmap = cmap)\n",
        "    plt.title(title)\n",
        "    plt.colorbar()\n",
        "    tick_marks = np.arange(len(classes))\n",
        "    plt.xticks(tick_marks, classes, rotation= 45)\n",
        "    plt.yticks(tick_marks, classes)\n",
        "    \n",
        "    if normalize:\n",
        "        cm = cm.astype('float') / cm.sum(axis = 1)[:, np.newaxis]\n",
        "        print(\"Normalized confusion matrix\")\n",
        "        \n",
        "    else:\n",
        "        print(\"confusion matrix not normalized\")\n",
        "    \n",
        "    \n",
        "    print(cm)\n",
        "    \n",
        "    thresh = cm.max() / 2\n",
        "    for i,j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
        "        plt.text(j, i, cm[i, j],\n",
        "                horizontalalignment = 'center',\n",
        "                color = 'white' if cm[i, j] > thresh else 'black')\n",
        "        \n",
        "        plt.tight_layout()\n",
        "        plt.ylabel('True Label')\n",
        "        plt.xlabel('Predicted Label')\n",
        "    "
      ],
      "execution_count": 113,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YAriWmHh0774",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "cm = confusion_matrix(test_labels, predictions.argmax(axis=1))"
      ],
      "execution_count": 114,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0uyXnKGn0-9a",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 484
        },
        "outputId": "ba9e9709-961d-47e6-a77f-9a3009064875"
      },
      "source": [
        "cm_plot_labels = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']\n",
        "plot_confusion_matrix(cm, cm_plot_labels, title = 'confusion_matrix')"
      ],
      "execution_count": 115,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "confusion matrix not normalized\n",
            "[[11  0  0  0  0  0  0  0  0  0]\n",
            " [ 0 11  0  0  0  0  0  0  0  0]\n",
            " [ 0  0 11  0  0  0  0  0  0  0]\n",
            " [ 0  0  0 11  0  0  0  0  0  0]\n",
            " [ 0  0  0  0 10  1  0  0  0  0]\n",
            " [ 0  0  0  0  0 11  0  0  0  0]\n",
            " [ 0  0  0  0  0  0 11  0  0  0]\n",
            " [ 0  0  0  0  0  0  0 11  0  0]\n",
            " [ 0  0  0  0  0  0  0  0 11  0]\n",
            " [ 0  0  0  0  0  0  0  0  0 11]]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}