banglalekha_numerals_cnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/mnpappo/anaconda2/lib/python2.7/site-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from __future__ import print_function\n",
    "import keras\n",
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential, Model\n",
    "from keras.layers import Dense, Dropout, Flatten, Activation, Input\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras import backend as K\n",
    "import numpy as np\n",
    "import gzip,cPickle,sys\n",
    "from matplotlib.pyplot import imshow\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "\n",
    "np.random.seed(1337)\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 32\n",
    "num_classes = 10\n",
    "epochs = 100\n",
    "img_width, img_height = 28, 28"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_data_dir = 'data/train/bn_numerals'\n",
    "validation_data_dir = 'data/validation/bn_numerals'\n",
    "nb_train_samples = 1560\n",
    "nb_validation_samples = 390"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "if K.image_data_format() == 'channels_first':\n",
    "    input_shape = (1, img_width, img_height)\n",
    "else:\n",
    "    input_shape = (img_width, img_height, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "img_input = Input(shape=input_shape)\n",
    "\n",
    "# Conv Block 1\n",
    "x = Conv2D(16, (3, 3), activation='relu', padding='same', name='block1_conv1')(img_input)\n",
    "x = Conv2D(32, (3, 3), activation='relu', padding='same', name='block1_conv2')(x)\n",
    "x = MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(x)\n",
    "\n",
    "# Conv Block 2\n",
    "x = Conv2D(32, (3, 3), activation='relu', padding='same', name='block2_conv1')(x)\n",
    "x = Conv2D(64, (3, 3), activation='relu', padding='same', name='block2_conv2')(x)\n",
    "x = MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(x)\n",
    "\n",
    "# Classification block\n",
    "x = Flatten(name='flatten')(x)\n",
    "x = Dense(32, activation='relu', name='fc')(x)\n",
    "x = Dense(num_classes, activation='softmax', name='predictions')(x)\n",
    "\n",
    "model = Model(img_input, x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 28, 28, 1)         0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 28, 28, 16)        160       \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 28, 28, 32)        4640      \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 14, 14, 32)        9248      \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 14, 14, 64)        18496     \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 3136)              0         \n",
      "_________________________________________________________________\n",
      "fc (Dense)                   (None, 32)                100384    \n",
      "_________________________________________________________________\n",
      "predictions (Dense)          (None, 10)                330       \n",
      "=================================================================\n",
      "Total params: 133,258\n",
      "Trainable params: 133,258\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# this is the augmentation configuration we will use for training\n",
    "train_datagen = ImageDataGenerator(\n",
    "    rescale=1./255,\n",
    "    rotation_range=0.5,\n",
    "    width_shift_range=0.2,\n",
    "    height_shift_range=0.2\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# only rescaling\n",
    "test_datagen = ImageDataGenerator(rescale=1./ 255)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 15600 images belonging to 10 classes.\n",
      "Found 4148 images belonging to 10 classes.\n"
     ]
    }
   ],
   "source": [
    "train_generator = train_datagen.flow_from_directory(\n",
    "    train_data_dir,\n",
    "    target_size=(img_width, img_height),\n",
    "    color_mode = \"grayscale\",\n",
    "    shuffle = True,\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical')\n",
    "\n",
    "validation_generator = test_datagen.flow_from_directory(\n",
    "    validation_data_dir,\n",
    "    target_size=(img_width, img_height),\n",
    "    color_mode = \"grayscale\",\n",
    "    shuffle = True,\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#tb = keras.callbacks.TensorBoard(log_dir='./logs', histogram_freq=0, write_graph=True, write_images=True, embeddings_freq=0, embeddings_layer_names=None, embeddings_metadata=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "48/48 [==============================] - 6s 134ms/step - loss: 1.8279 - acc: 0.3568 - val_loss: 0.9017 - val_acc: 0.7396\n",
      "Epoch 2/100\n",
      "48/48 [==============================] - 5s 108ms/step - loss: 1.0835 - acc: 0.6341 - val_loss: 0.4724 - val_acc: 0.8438\n",
      "Epoch 3/100\n",
      "48/48 [==============================] - 5s 114ms/step - loss: 0.7791 - acc: 0.7428 - val_loss: 0.4314 - val_acc: 0.8516\n",
      "Epoch 4/100\n",
      "48/48 [==============================] - 6s 126ms/step - loss: 0.6977 - acc: 0.7689 - val_loss: 0.4077 - val_acc: 0.8620\n",
      "Epoch 5/100\n",
      "48/48 [==============================] - 6s 124ms/step - loss: 0.6579 - acc: 0.7943 - val_loss: 0.3206 - val_acc: 0.8932\n",
      "Epoch 6/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.5542 - acc: 0.8125 - val_loss: 0.4017 - val_acc: 0.8542\n",
      "Epoch 7/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.5217 - acc: 0.8288 - val_loss: 0.3547 - val_acc: 0.8984\n",
      "Epoch 8/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.4599 - acc: 0.8503 - val_loss: 0.2311 - val_acc: 0.9427\n",
      "Epoch 9/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.4250 - acc: 0.8691 - val_loss: 0.2447 - val_acc: 0.9271\n",
      "Epoch 10/100\n",
      "48/48 [==============================] - 6s 122ms/step - loss: 0.4376 - acc: 0.8607 - val_loss: 0.2445 - val_acc: 0.9349\n",
      "Epoch 11/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.4130 - acc: 0.8757 - val_loss: 0.1803 - val_acc: 0.9328\n",
      "Epoch 12/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.4244 - acc: 0.8711 - val_loss: 0.2709 - val_acc: 0.9219\n",
      "Epoch 13/100\n",
      "48/48 [==============================] - 6s 121ms/step - loss: 0.4452 - acc: 0.8555 - val_loss: 0.2545 - val_acc: 0.9245\n",
      "Epoch 14/100\n",
      "48/48 [==============================] - 6s 124ms/step - loss: 0.3297 - acc: 0.8828 - val_loss: 0.1820 - val_acc: 0.9479\n",
      "Epoch 15/100\n",
      "48/48 [==============================] - 6s 122ms/step - loss: 0.3771 - acc: 0.8796 - val_loss: 0.1847 - val_acc: 0.9427\n",
      "Epoch 16/100\n",
      "48/48 [==============================] - 6s 121ms/step - loss: 0.3279 - acc: 0.8997 - val_loss: 0.2747 - val_acc: 0.9089\n",
      "Epoch 17/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.3573 - acc: 0.8867 - val_loss: 0.2670 - val_acc: 0.9193\n",
      "Epoch 18/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.3312 - acc: 0.8919 - val_loss: 0.1699 - val_acc: 0.9453\n",
      "Epoch 19/100\n",
      "48/48 [==============================] - 6s 129ms/step - loss: 0.3062 - acc: 0.9062 - val_loss: 0.1469 - val_acc: 0.9635\n",
      "Epoch 20/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.3168 - acc: 0.9023 - val_loss: 0.2504 - val_acc: 0.9349\n",
      "Epoch 21/100\n",
      "48/48 [==============================] - 6s 125ms/step - loss: 0.2528 - acc: 0.9154 - val_loss: 0.2236 - val_acc: 0.9297\n",
      "Epoch 22/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.2731 - acc: 0.9076 - val_loss: 0.1762 - val_acc: 0.9435\n",
      "Epoch 23/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.2870 - acc: 0.9082 - val_loss: 0.1298 - val_acc: 0.9557\n",
      "Epoch 24/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.2960 - acc: 0.9030 - val_loss: 0.2167 - val_acc: 0.9401\n",
      "Epoch 25/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.2760 - acc: 0.9134 - val_loss: 0.1019 - val_acc: 0.9766\n",
      "Epoch 26/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.2772 - acc: 0.8958 - val_loss: 0.2076 - val_acc: 0.9531\n",
      "Epoch 27/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.3001 - acc: 0.9076 - val_loss: 0.1593 - val_acc: 0.9583\n",
      "Epoch 28/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.2421 - acc: 0.9271 - val_loss: 0.1916 - val_acc: 0.9427\n",
      "Epoch 29/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.2257 - acc: 0.9232 - val_loss: 0.2034 - val_acc: 0.9401\n",
      "Epoch 30/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.2344 - acc: 0.9251 - val_loss: 0.1757 - val_acc: 0.9583\n",
      "Epoch 31/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.2392 - acc: 0.9186 - val_loss: 0.2167 - val_acc: 0.9323\n",
      "Epoch 32/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.2690 - acc: 0.9082 - val_loss: 0.2019 - val_acc: 0.9427\n",
      "Epoch 33/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.2023 - acc: 0.9368 - val_loss: 0.0998 - val_acc: 0.9731\n",
      "Epoch 34/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.2139 - acc: 0.9329 - val_loss: 0.1174 - val_acc: 0.9714\n",
      "Epoch 35/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.2637 - acc: 0.9206 - val_loss: 0.2312 - val_acc: 0.9401\n",
      "Epoch 36/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.2450 - acc: 0.9271 - val_loss: 0.1572 - val_acc: 0.9688\n",
      "Epoch 37/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.2284 - acc: 0.9206 - val_loss: 0.1662 - val_acc: 0.9609\n",
      "Epoch 38/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.2125 - acc: 0.9297 - val_loss: 0.1451 - val_acc: 0.9453\n",
      "Epoch 39/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.1903 - acc: 0.9349 - val_loss: 0.2343 - val_acc: 0.9401\n",
      "Epoch 40/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.2262 - acc: 0.9290 - val_loss: 0.2191 - val_acc: 0.9635\n",
      "Epoch 41/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.1972 - acc: 0.9375 - val_loss: 0.1420 - val_acc: 0.9531\n",
      "Epoch 42/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1997 - acc: 0.9388 - val_loss: 0.1064 - val_acc: 0.9583\n",
      "Epoch 43/100\n",
      "48/48 [==============================] - 6s 116ms/step - loss: 0.1923 - acc: 0.9421 - val_loss: 0.1628 - val_acc: 0.9557\n",
      "Epoch 44/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.1806 - acc: 0.9408 - val_loss: 0.1273 - val_acc: 0.9624\n",
      "Epoch 45/100\n",
      "48/48 [==============================] - 5s 114ms/step - loss: 0.1925 - acc: 0.9362 - val_loss: 0.0984 - val_acc: 0.9635\n",
      "Epoch 46/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.1989 - acc: 0.9368 - val_loss: 0.1079 - val_acc: 0.9766\n",
      "Epoch 47/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1797 - acc: 0.9479 - val_loss: 0.1631 - val_acc: 0.9583\n",
      "Epoch 48/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.1872 - acc: 0.9421 - val_loss: 0.1002 - val_acc: 0.9688\n",
      "Epoch 49/100\n",
      "48/48 [==============================] - 6s 124ms/step - loss: 0.2113 - acc: 0.9342 - val_loss: 0.1988 - val_acc: 0.9479\n",
      "Epoch 50/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1790 - acc: 0.9395 - val_loss: 0.2468 - val_acc: 0.9401\n",
      "Epoch 51/100\n",
      "48/48 [==============================] - 6s 116ms/step - loss: 0.2022 - acc: 0.9427 - val_loss: 0.1442 - val_acc: 0.9635\n",
      "Epoch 52/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1768 - acc: 0.9401 - val_loss: 0.1075 - val_acc: 0.9714\n",
      "Epoch 53/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.1816 - acc: 0.9486 - val_loss: 0.1056 - val_acc: 0.9583\n",
      "Epoch 54/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.1894 - acc: 0.9375 - val_loss: 0.1834 - val_acc: 0.9297\n",
      "Epoch 55/100\n",
      "48/48 [==============================] - 6s 122ms/step - loss: 0.1840 - acc: 0.9401 - val_loss: 0.1113 - val_acc: 0.9677\n",
      "Epoch 56/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1878 - acc: 0.9453 - val_loss: 0.1570 - val_acc: 0.9583\n",
      "Epoch 57/100\n",
      "48/48 [==============================] - 6s 116ms/step - loss: 0.1684 - acc: 0.9473 - val_loss: 0.1022 - val_acc: 0.9714\n",
      "Epoch 58/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.1736 - acc: 0.9388 - val_loss: 0.1583 - val_acc: 0.9557\n",
      "Epoch 59/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1384 - acc: 0.9551 - val_loss: 0.1139 - val_acc: 0.9688\n",
      "Epoch 60/100\n",
      "48/48 [==============================] - 6s 128ms/step - loss: 0.1574 - acc: 0.9440 - val_loss: 0.1001 - val_acc: 0.9714\n",
      "Epoch 61/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1639 - acc: 0.9453 - val_loss: 0.1499 - val_acc: 0.9688\n",
      "Epoch 62/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1707 - acc: 0.9473 - val_loss: 0.1066 - val_acc: 0.9714\n",
      "Epoch 63/100\n",
      "48/48 [==============================] - 6s 120ms/step - loss: 0.1764 - acc: 0.9401 - val_loss: 0.0955 - val_acc: 0.9792\n",
      "Epoch 64/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1637 - acc: 0.9473 - val_loss: 0.1260 - val_acc: 0.9688\n",
      "Epoch 65/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1385 - acc: 0.9531 - val_loss: 0.1642 - val_acc: 0.9462\n",
      "Epoch 66/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1697 - acc: 0.9538 - val_loss: 0.0881 - val_acc: 0.9740\n",
      "Epoch 67/100\n",
      "48/48 [==============================] - 6s 122ms/step - loss: 0.1839 - acc: 0.9434 - val_loss: 0.0844 - val_acc: 0.9766\n",
      "Epoch 68/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.1527 - acc: 0.9570 - val_loss: 0.0819 - val_acc: 0.9818\n",
      "Epoch 69/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.1574 - acc: 0.9460 - val_loss: 0.1308 - val_acc: 0.9661\n",
      "Epoch 70/100\n",
      "48/48 [==============================] - 6s 121ms/step - loss: 0.1668 - acc: 0.9531 - val_loss: 0.2053 - val_acc: 0.9375\n",
      "Epoch 71/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1376 - acc: 0.9544 - val_loss: 0.0485 - val_acc: 0.9922\n",
      "Epoch 72/100\n",
      "48/48 [==============================] - 6s 125ms/step - loss: 0.1537 - acc: 0.9499 - val_loss: 0.0827 - val_acc: 0.9844\n",
      "Epoch 73/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1737 - acc: 0.9447 - val_loss: 0.1153 - val_acc: 0.9661\n",
      "Epoch 74/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1508 - acc: 0.9505 - val_loss: 0.1099 - val_acc: 0.9688\n",
      "Epoch 75/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1460 - acc: 0.9551 - val_loss: 0.1023 - val_acc: 0.9661\n",
      "Epoch 76/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1453 - acc: 0.9525 - val_loss: 0.1898 - val_acc: 0.9624\n",
      "Epoch 77/100\n",
      "48/48 [==============================] - 6s 116ms/step - loss: 0.1462 - acc: 0.9525 - val_loss: 0.1385 - val_acc: 0.9531\n",
      "Epoch 78/100\n",
      "48/48 [==============================] - 5s 115ms/step - loss: 0.1613 - acc: 0.9486 - val_loss: 0.1013 - val_acc: 0.9661\n",
      "Epoch 79/100\n",
      "48/48 [==============================] - 6s 121ms/step - loss: 0.1527 - acc: 0.9538 - val_loss: 0.0912 - val_acc: 0.9714\n",
      "Epoch 80/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1415 - acc: 0.9544 - val_loss: 0.0941 - val_acc: 0.9714\n",
      "Epoch 81/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1500 - acc: 0.9551 - val_loss: 0.1214 - val_acc: 0.9688\n",
      "Epoch 82/100\n",
      "48/48 [==============================] - 5s 113ms/step - loss: 0.1407 - acc: 0.9512 - val_loss: 0.1069 - val_acc: 0.9766\n",
      "Epoch 83/100\n",
      "48/48 [==============================] - 6s 119ms/step - loss: 0.1331 - acc: 0.9583 - val_loss: 0.1043 - val_acc: 0.9714\n",
      "Epoch 84/100\n",
      "48/48 [==============================] - 5s 113ms/step - loss: 0.1333 - acc: 0.9564 - val_loss: 0.1032 - val_acc: 0.9740\n",
      "Epoch 85/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1426 - acc: 0.9590 - val_loss: 0.1611 - val_acc: 0.9635\n",
      "Epoch 86/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1689 - acc: 0.9447 - val_loss: 0.0980 - val_acc: 0.9792\n",
      "Epoch 87/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1510 - acc: 0.9564 - val_loss: 0.1618 - val_acc: 0.9435\n",
      "Epoch 88/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1460 - acc: 0.9531 - val_loss: 0.1022 - val_acc: 0.9766\n",
      "Epoch 89/100\n",
      "48/48 [==============================] - 6s 116ms/step - loss: 0.1337 - acc: 0.9551 - val_loss: 0.1028 - val_acc: 0.9635\n",
      "Epoch 90/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.1341 - acc: 0.9616 - val_loss: 0.1476 - val_acc: 0.9583\n",
      "Epoch 91/100\n",
      "48/48 [==============================] - 6s 116ms/step - loss: 0.1554 - acc: 0.9512 - val_loss: 0.1214 - val_acc: 0.9661\n",
      "Epoch 92/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1226 - acc: 0.9616 - val_loss: 0.0969 - val_acc: 0.9688\n",
      "Epoch 93/100\n",
      "48/48 [==============================] - 6s 125ms/step - loss: 0.1285 - acc: 0.9570 - val_loss: 0.1233 - val_acc: 0.9661\n",
      "Epoch 94/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1230 - acc: 0.9642 - val_loss: 0.0561 - val_acc: 0.9870\n",
      "Epoch 95/100\n",
      "48/48 [==============================] - 6s 115ms/step - loss: 0.1416 - acc: 0.9551 - val_loss: 0.1006 - val_acc: 0.9766\n",
      "Epoch 96/100\n",
      "48/48 [==============================] - 6s 116ms/step - loss: 0.1456 - acc: 0.9518 - val_loss: 0.1631 - val_acc: 0.9661\n",
      "Epoch 97/100\n",
      "48/48 [==============================] - 6s 122ms/step - loss: 0.1543 - acc: 0.9518 - val_loss: 0.1730 - val_acc: 0.9557\n",
      "Epoch 98/100\n",
      "48/48 [==============================] - 6s 118ms/step - loss: 0.1359 - acc: 0.9551 - val_loss: 0.1141 - val_acc: 0.9785\n",
      "Epoch 99/100\n",
      "48/48 [==============================] - 6s 117ms/step - loss: 0.1190 - acc: 0.9590 - val_loss: 0.0743 - val_acc: 0.9766\n",
      "Epoch 100/100\n",
      "48/48 [==============================] - 6s 123ms/step - loss: 0.1315 - acc: 0.9570 - val_loss: 0.1275 - val_acc: 0.9635\n"
     ]
    }
   ],
   "source": [
    "history = model.fit_generator(\n",
    "    train_generator,\n",
    "    steps_per_epoch=nb_train_samples // batch_size,\n",
    "    epochs=epochs,\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=nb_validation_samples // batch_size,\n",
    "    #callbacks=[tb],\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1828673ad0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEWCAYAAAB1xKBvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nbQ8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1828298d10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# accuracy\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('conv1d model accuracy - short data')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "#loss\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('conv1d model loss - short data')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#saving model\n",
    "#model.save_weights('bn_digit.h5')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}