Telugu_script_digits.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Tc4FoEagMzHg"
   },
   "source": [
    "## Telugu Script - Digits Recognition"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "IZ8S-VLcvWil"
   },
   "source": [
    "https://en.wikipedia.org/wiki/Telugu_script"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "ct-HdffLMyql"
   },
   "source": [
    "Telugu script, an abugida from the Brahmic family of scripts, is used to write the Telugu language, a Dravidian language spoken in the Indian states of Andhra Pradesh and Telangana as well as several other neighbouring states."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "YaKAKU7yBWid",
    "outputId": "3181bba5-a3ef-4cef-ae86-6718e70d86bc"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "vAqnShVRAb8q",
    "outputId": "b526e63e-2ef5-4bb4-9586-0c6bf8f6e9b3"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "# Importing all necessary libraries \n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras.models import Sequential,Model\n",
    "from keras.utils.np_utils import to_categorical\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Flatten,BatchNormalization,Input\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras import backend as K\n",
    "from keras.layers import Activation, Dense\n",
    "from keras import optimizers\n",
    "from keras.layers import Flatten\n",
    "import tensorflow as tf\n",
    "import glob\n",
    "from PIL import Image\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from numpy import expand_dims"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "cGqW47m9Ab85"
   },
   "outputs": [],
   "source": [
    "# get images from drive\n",
    "images = glob.glob(\"/content/drive/My Drive/DLCP/telugu_script/numerals\"+\"/*.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_MJWIr7NAb9A"
   },
   "outputs": [],
   "source": [
    "# sort images \n",
    "images.sort()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 187
    },
    "colab_type": "code",
    "id": "cXssBiwLAb9E",
    "outputId": "6c7ec4c7-022e-432e-c2f3-efeabc31898b"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['/content/drive/My Drive/DLCP/telugu_script/numerals/0.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/1.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/2.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/3.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/4.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/5.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/6.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/7.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/8.png',\n",
       " '/content/drive/My Drive/DLCP/telugu_script/numerals/9.png']"
      ]
     },
     "execution_count": 6,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "-keGeXEvAb9I"
   },
   "outputs": [],
   "source": [
    "# convert images to grayscale and resize to 28X28\n",
    "images_array = []\n",
    "for img in images:\n",
    "    img = Image.open(img)\n",
    "    # img = img.convert('RGB')\n",
    "    img = img.convert('L')\n",
    "    img = img.resize((28,28), Image.ANTIALIAS)\n",
    "    images_array.append(np.array(img))\n",
    "images_array = np.array(images_array)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 98
    },
    "colab_type": "code",
    "id": "-DnF0GlQAb9V",
    "outputId": "ee7d4f62-a5cf-4826-cdb8-c53b077d420f"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlIAAABRCAYAAAAZ1Ej0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<Figure size 720x72 with 10 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot images\n",
    "plt.figure(figsize=(10, 1))\n",
    "for i in range(10):\n",
    "    plt.subplot(1,10,i+1)\n",
    "    plt.imshow(images_array[i],cmap=\"gray\")\n",
    "    plt.axis(\"off\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "4Y51ecoYAb9c"
   },
   "outputs": [],
   "source": [
    "# create a large dataset with duplicates \n",
    "final_array = []\n",
    "labels_array = []\n",
    "for i,img in enumerate(images_array):\n",
    "    for j in range(10000):\n",
    "        final_array.append(img)\n",
    "        labels_array.append(i)\n",
    "final_array = np.array(final_array)\n",
    "labels_array = np.array(labels_array)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "jZhY6oJvAb9t"
   },
   "outputs": [],
   "source": [
    "# split the dataset into train and test\n",
    "trainX,testX,trainy,testy = train_test_split(final_array,labels_array,shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "R5icTKlJAb9x"
   },
   "outputs": [],
   "source": [
    "# reshape to be [samples][width][height][channels]\n",
    "X_train = trainX.reshape((trainX.shape[0], 28, 28, 1)).astype('float32')\n",
    "X_test = testX.reshape((testX.shape[0], 28, 28, 1)).astype('float32')\n",
    "# X_train = trainX.reshape(-1,28,28,1)\n",
    "# X_test = testX.reshape(-1,28,28,1)\n",
    "\n",
    "# normalize inputs from 0-255 to 0-1\n",
    "X_train = X_train / 255\n",
    "X_test = X_test / 255\n",
    "\n",
    "# convert\n",
    "Y_train = tf.keras.utils.to_categorical(trainy, num_classes=10)\n",
    "Y_test = tf.keras.utils.to_categorical(testy, num_classes=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "aijVAzC8UmJg",
    "outputId": "71f5fe80-cfe3-4142-f59d-0140762670be"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(75000, 28, 28, 1) (25000, 28, 28, 1) (75000, 10) (25000, 10)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_test.shape, Y_train.shape, Y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "gugWXryEck8V"
   },
   "source": [
    "## Image agumentation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 71
    },
    "colab_type": "code",
    "id": "Ne_x6jq8aFCi",
    "outputId": "4aae09f7-05dd-48c8-9c2b-a0c4957dcbed"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/keras_preprocessing/image/image_data_generator.py:341: UserWarning: This ImageDataGenerator specifies `zca_whitening` which overrides setting of`featurewise_std_normalization`.\n",
      "  warnings.warn('This ImageDataGenerator specifies '\n"
     ]
    }
   ],
   "source": [
    "train_image_gen = ImageDataGenerator(\n",
    "    featurewise_center=True,\n",
    "    featurewise_std_normalization=True,\n",
    "    zca_whitening=True,\n",
    "    rotation_range=90,\n",
    "    horizontal_flip=True, \n",
    "    vertical_flip=True,\n",
    "    # rescale=1./255,\n",
    "    width_shift_range=0.2,\n",
    "    height_shift_range=0.2)\n",
    "test_image_gen = ImageDataGenerator(\n",
    "    featurewise_center=True,\n",
    "    featurewise_std_normalization=True,\n",
    "    zca_whitening=True,\n",
    "    rotation_range=90,\n",
    "    horizontal_flip=True, \n",
    "    vertical_flip=True,\n",
    "    # rescale=1./255,\n",
    "    width_shift_range=0.2,\n",
    "    height_shift_range=0.2)\n",
    "\n",
    "train_image_gen.fit(X_train, augment=True)\n",
    "test_image_gen.fit(X_test, augment=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 269
    },
    "colab_type": "code",
    "id": "tGq4CYq0Ab-F",
    "outputId": "0c155baf-3a2b-44d9-954c-b92be3e2a95c"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVEAAAD8CAYAAADOg5fGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 9 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot images\n",
    "for X_batch, y_batch in train_image_gen.flow(X_train, trainy, batch_size=9):\n",
    "    # create a grid of 3x3 images\n",
    "    for i in range(0, 9):\n",
    "        plt.subplot(330 + 1 + i)\n",
    "        plt.imshow(X_batch[i].reshape(28, 28), cmap=plt.get_cmap('gray'))\n",
    "    # show the plot\n",
    "    plt.show()\n",
    "    break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0aSxstldAb-K"
   },
   "outputs": [],
   "source": [
    "# lets set epochs and batch_size values\n",
    "epochs = 10\n",
    "batch_size = 128"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "tsaKyciRbTcV"
   },
   "source": [
    "## CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 734
    },
    "colab_type": "code",
    "id": "btzbX01AoKtm",
    "outputId": "c3567f2d-7dbf-415b-a177-da4bb2d39184"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/init_ops.py:1251: calling VarianceScaling.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Call initializer instance with the dtype argument instead of passing it to the constructor\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 27, 27, 128)       640       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 13, 13, 128)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 13, 13, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 12, 12, 64)        32832     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 6, 6, 64)          256       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 5, 5, 32)          8224      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 2, 2, 32)          0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 2, 2, 32)          128       \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 32)                128       \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 53,386\n",
      "Trainable params: 52,874\n",
      "Non-trainable params: 512\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = tf.keras.models.Sequential()\n",
    "# model.add(tf.keras.layers.Reshape((784,),input_shape=(28,28)))\n",
    "# model.add(tf.keras.layers.Reshape((28,28,1),input_shape=(28,28)))\n",
    "model.add(tf.keras.layers.Conv2D(128,kernel_size=(2,2), input_shape =(28,28,1), activation=\"relu\"))\n",
    "model.add(tf.keras.layers.MaxPooling2D(pool_size =(2,2)))\n",
    "model.add(tf.keras.layers.BatchNormalization())\n",
    "\n",
    "model.add(tf.keras.layers.Conv2D(64,kernel_size=(2,2), activation=\"relu\"))\n",
    "model.add(tf.keras.layers.MaxPooling2D(pool_size =(2,2)))\n",
    "model.add(tf.keras.layers.BatchNormalization())\n",
    "\n",
    "model.add(tf.keras.layers.Conv2D(32,kernel_size=(2,2), activation=\"relu\"))\n",
    "model.add(tf.keras.layers.MaxPooling2D(pool_size =(2,2)))\n",
    "model.add(tf.keras.layers.BatchNormalization())\n",
    "\n",
    "model.add(tf.keras.layers.Flatten())\n",
    "model.add(tf.keras.layers.Dense(64, activation=\"relu\"))\n",
    "model.add(tf.keras.layers.Dropout(0.2))\n",
    "model.add(tf.keras.layers.Dense(32,kernel_regularizer = tf.keras.regularizers.l2(0.01), activation=\"relu\"))\n",
    "model.add(tf.keras.layers.BatchNormalization())\n",
    "model.add(tf.keras.layers.Dense(10,activation=\"softmax\"))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ap2roeumAb-Q"
   },
   "outputs": [],
   "source": [
    "# Build CNN\n",
    "model.compile(loss='categorical_crossentropy', optimizer='sgd', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "K8ZHj52DLkzZ",
    "outputId": "4a534cd5-b64a-4984-da40-225cde5eece0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 1.8902 - acc: 0.4576\n",
      "Epoch 00001: val_acc improved from -inf to 0.73080, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 84s 144ms/step - loss: 1.8888 - acc: 0.4581 - val_loss: 1.1407 - val_acc: 0.7308\n",
      "Epoch 2/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.9467 - acc: 0.7655\n",
      "Epoch 00002: val_acc improved from 0.73080 to 0.85260, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 80s 136ms/step - loss: 0.9464 - acc: 0.7656 - val_loss: 0.6694 - val_acc: 0.8526\n",
      "Epoch 3/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.6569 - acc: 0.8352\n",
      "Epoch 00003: val_acc improved from 0.85260 to 0.87732, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 79s 135ms/step - loss: 0.6567 - acc: 0.8353 - val_loss: 0.5111 - val_acc: 0.8773\n",
      "Epoch 4/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.5137 - acc: 0.8657\n",
      "Epoch 00004: val_acc improved from 0.87732 to 0.90108, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 77s 131ms/step - loss: 0.5135 - acc: 0.8658 - val_loss: 0.4080 - val_acc: 0.9011\n",
      "Epoch 5/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.4264 - acc: 0.8831\n",
      "Epoch 00005: val_acc improved from 0.90108 to 0.90828, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 76s 130ms/step - loss: 0.4265 - acc: 0.8831 - val_loss: 0.3498 - val_acc: 0.9083\n",
      "Epoch 6/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.3585 - acc: 0.8999\n",
      "Epoch 00006: val_acc improved from 0.90828 to 0.92424, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 77s 131ms/step - loss: 0.3585 - acc: 0.8999 - val_loss: 0.2901 - val_acc: 0.9242\n",
      "Epoch 7/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.3090 - acc: 0.9116\n",
      "Epoch 00007: val_acc improved from 0.92424 to 0.93600, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 75s 129ms/step - loss: 0.3089 - acc: 0.9116 - val_loss: 0.2468 - val_acc: 0.9360\n",
      "Epoch 8/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.2658 - acc: 0.9228\n",
      "Epoch 00008: val_acc improved from 0.93600 to 0.94628, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 76s 129ms/step - loss: 0.2657 - acc: 0.9229 - val_loss: 0.2097 - val_acc: 0.9463\n",
      "Epoch 9/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.2318 - acc: 0.9331\n",
      "Epoch 00009: val_acc improved from 0.94628 to 0.95272, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 77s 131ms/step - loss: 0.2318 - acc: 0.9331 - val_loss: 0.1815 - val_acc: 0.9527\n",
      "Epoch 10/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.2050 - acc: 0.9397\n",
      "Epoch 00010: val_acc improved from 0.95272 to 0.95564, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 75s 128ms/step - loss: 0.2050 - acc: 0.9397 - val_loss: 0.1594 - val_acc: 0.9556\n",
      "Epoch 11/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.1787 - acc: 0.9474\n",
      "Epoch 00011: val_acc did not improve from 0.95564\n",
      "586/585 [==============================] - 76s 129ms/step - loss: 0.1786 - acc: 0.9475 - val_loss: 0.1499 - val_acc: 0.9542\n",
      "Epoch 12/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.1614 - acc: 0.9511\n",
      "Epoch 00012: val_acc improved from 0.95564 to 0.95736, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 76s 130ms/step - loss: 0.1615 - acc: 0.9511 - val_loss: 0.1437 - val_acc: 0.9574\n",
      "Epoch 13/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.1450 - acc: 0.9549\n",
      "Epoch 00013: val_acc improved from 0.95736 to 0.96692, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 89s 152ms/step - loss: 0.1449 - acc: 0.9549 - val_loss: 0.1125 - val_acc: 0.9669\n",
      "Epoch 14/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.1313 - acc: 0.9585\n",
      "Epoch 00014: val_acc improved from 0.96692 to 0.96868, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 76s 129ms/step - loss: 0.1313 - acc: 0.9584 - val_loss: 0.1010 - val_acc: 0.9687\n",
      "Epoch 15/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.1224 - acc: 0.9611\n",
      "Epoch 00015: val_acc improved from 0.96868 to 0.97044, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 75s 128ms/step - loss: 0.1225 - acc: 0.9611 - val_loss: 0.0947 - val_acc: 0.9704\n",
      "Epoch 16/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.1126 - acc: 0.9634\n",
      "Epoch 00016: val_acc improved from 0.97044 to 0.97560, saving model to /content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\n",
      "586/585 [==============================] - 75s 128ms/step - loss: 0.1126 - acc: 0.9634 - val_loss: 0.0809 - val_acc: 0.9756\n",
      "Epoch 17/30\n",
      "585/585 [============================>.] - ETA: 0s - loss: 0.1054 - acc: 0.9653\n",
      "Epoch 00017: val_acc did not improve from 0.97560\n",
      "586/585 [==============================] - 75s 127ms/step - loss: 0.1053 - acc: 0.9653 - val_loss: 0.0916 - val_acc: 0.9705\n",
      "Epoch 00017: early stopping\n"
     ]
    }
   ],
   "source": [
    "checkpint = tf.keras.callbacks.ModelCheckpoint(filepath=\"/content/drive/My Drive/DLCP/telugu_script/tdigits_model.h5\",save_best_only=True,verbose=1,monitor=\"val_acc\",mode=\"max\")\n",
    "stoperly = tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\",mode=\"auto\",verbose=1)\n",
    "callbaklist = [checkpint,stoperly]\n",
    "# model_history = model.fit(X_train_m3,trainy,validation_data=(X_test_m3,testy),epochs=10,batch_size=128,callbacks=callbaklist)\n",
    "model_history = model.fit_generator(train_image_gen.flow(X_train, Y_train, batch_size=batch_size),validation_data=test_image_gen.flow(X_test, Y_test, batch_size=batch_size),\n",
    "                    steps_per_epoch=len(X_train) / batch_size,validation_steps=len(X_test)/batch_size, epochs = 30,verbose = 1,callbacks=callbaklist)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 503
    },
    "colab_type": "code",
    "id": "wZJZDFWk0YKo",
    "outputId": "d26d9782-4e84-4cae-bc66-6a6bde14baa7"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f45f028ca20>"
      ]
     },
     "execution_count": 19,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe0AAAHVCAYAAADcnaM7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFhERSREKbRERkRSh0BYREUkR/x9rcU7s2B+XtwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot train and test loss\n",
    "fig, ax = plt.subplots(figsize=(8, 8))\n",
    "ax.plot(model_history.history[\"loss\"])\n",
    "ax.legend(\"loss\")\n",
    "ax.plot(model_history.history[\"val_loss\"])\n",
    "ax.legend(\"val_loss\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 503
    },
    "colab_type": "code",
    "id": "1MsmpdES0eC-",
    "outputId": "e256903d-8edc-4eff-81d9-70c60e608602"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f45f89b2ac8>"
      ]
     },
     "execution_count": 20,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAecAAAHVCAYAAADLvzPyAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAc7SS+irIxLBAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot train and test accuracy\n",
    "fig, ax = plt.subplots(figsize=(8, 8))\n",
    "ax.plot(model_history.history[\"acc\"])\n",
    "ax.legend(\"acc\")\n",
    "ax.plot(model_history.history[\"val_acc\"])\n",
    "ax.legend(\"val_acc\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "zwM4CGFhc9NX",
    "outputId": "c05fcc8e-5393-4518-f0ab-6a26b5554d4e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.08662786245422095\n",
      "Test accuracy: 0.97176\n"
     ]
    }
   ],
   "source": [
    "# predict test accuracy\n",
    "scores = model.evaluate(test_image_gen.flow(X_test, Y_test, batch_size=batch_size), verbose=0)\n",
    "print('Test loss:', scores[0])\n",
    "print('Test accuracy:', scores[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0NIggEPPYnxJ"
   },
   "outputs": [],
   "source": [
    "# make a prediction using test-time augmentation\n",
    "def tta_prediction(datagen, model, image, n_examples):\n",
    "\t# convert image into dataset\n",
    "\tsamples = expand_dims(image, 0)\n",
    "\t# prepare iterator\n",
    "\tit = datagen.flow(samples, batch_size=n_examples)\n",
    "\t# make predictions for each augmented image\n",
    "\tyhats = model.predict_generator(it, steps=n_examples, verbose=0)\n",
    "\t# sum across predictions\n",
    "\tsummed = np.sum(yhats, axis=0)\n",
    "\t# argmax across classes\n",
    "\treturn np.argmax(summed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 357
    },
    "colab_type": "code",
    "id": "KjKDlgRCdGio",
    "outputId": "fea3ea97-ae67-46c9-8ff9-2f5e312d486a"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "truth: 0\n",
      "prediction: 0\n",
      "truth: 8\n",
      "prediction: 8\n",
      "truth: 1\n",
      "prediction: 1\n",
      "truth: 2\n",
      "prediction: 2\n",
      "truth: 0\n",
      "prediction: 0\n",
      "truth: 5\n",
      "prediction: 5\n",
      "truth: 0\n",
      "prediction: 0\n",
      "truth: 9\n",
      "prediction: 9\n",
      "truth: 3\n",
      "prediction: 3\n",
      "truth: 3\n",
      "prediction: 3\n"
     ]
    }
   ],
   "source": [
    "# plot and print model prediciton \n",
    "#plt.figure(figsize=(10, 1))\n",
    "for index in np.random.choice(25000,10):\n",
    "    img = X_test[index]\n",
    "    # plt.subplot(1,10,index+1)\n",
    "    # plt.imshow(img,cmap=\"gray\")\n",
    "    # plt.axis(\"off\")\n",
    "    print(\"truth:\",np.argmax(Y_test[index]))\n",
    "    pred = tta_prediction(train_image_gen,model,img,7)\n",
    "    print(\"prediction:\",pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vdJD1O6uKXWO"
   },
   "outputs": [],
   "source": [
    "# from keras.models import load_model\n",
    "# model = load_model('model.h5')\n",
    "# model.load_weights('/content/drive/My Drive/DLCP/model.h5')\n",
    "# pred = model.predict(X_test_m3[467].reshape(1,28,28,1))\n",
    "\n",
    "# plt.figure(figsize=(10, 1))\n",
    "# for index in np.random.choice(25000,10):\n",
    "#     img = Image.open('data/' + str(index) + '.png').convert(\"L\")\n",
    "#     img = img.resize((28,28))\n",
    "#     im2arr = np.array(img)\n",
    "#     im2arr = X_test_m3[index]\n",
    "#     im2arr = im2arr.reshape(1,28,28,1)\n",
    "#     Predicting the Test set results\n",
    "#     y_pred = model.predict(im2arr,batch_size=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DzKSfYgE5VWG"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "telugu_script_digits.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}