Traffic Sign Classification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using Theano backend.\n",
      "Using gpu device 0: GeForce GTX TITAN X (CNMeM is disabled, cuDNN 4007)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from skimage import io, color, exposure, transform\n",
    "from sklearn.cross_validation import train_test_split\n",
    "import os\n",
    "import glob\n",
    "import h5py\n",
    "\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential, model_from_json\n",
    "from keras.layers.core import Dense, Dropout, Activation, Flatten\n",
    "from keras.layers.convolutional import Conv2D\n",
    "from keras.layers.pooling import MaxPooling2D\n",
    "\n",
    "from keras.optimizers import SGD\n",
    "from keras.utils import np_utils\n",
    "from keras.callbacks import LearningRateScheduler, ModelCheckpoint\n",
    "from keras import backend as K\n",
    "K.set_image_data_format('channels_first')\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "NUM_CLASSES = 43\n",
    "IMG_SIZE = 48"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Function to preprocess the image:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def preprocess_img(img):\n",
    "    # Histogram normalization in y\n",
    "    hsv = color.rgb2hsv(img)\n",
    "    hsv[:,:,2] = exposure.equalize_hist(hsv[:,:,2])\n",
    "    img = color.hsv2rgb(hsv)\n",
    "\n",
    "    # central scrop\n",
    "    min_side = min(img.shape[:-1])\n",
    "    centre = img.shape[0]//2, img.shape[1]//2\n",
    "    img = img[centre[0]-min_side//2:centre[0]+min_side//2,\n",
    "              centre[1]-min_side//2:centre[1]+min_side//2,\n",
    "              :]\n",
    "\n",
    "    # rescale to standard size\n",
    "    img = transform.resize(img, (IMG_SIZE, IMG_SIZE))\n",
    "\n",
    "    # roll color axis to axis 0\n",
    "    img = np.rollaxis(img,-1)\n",
    "\n",
    "    return img\n",
    "\n",
    "\n",
    "def get_class(img_path):\n",
    "    return int(img_path.split('/')[-2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "## Preprocess all training images into a numpy array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loaded images from X.h5\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    with  h5py.File('X.h5') as hf: \n",
    "        X, Y = hf['imgs'][:], hf['labels'][:]\n",
    "    print(\"Loaded images from X.h5\")\n",
    "    \n",
    "except (IOError,OSError, KeyError):  \n",
    "    print(\"Error in reading X.h5. Processing all images...\")\n",
    "    root_dir = 'GTSRB/Final_Training/Images/'\n",
    "    imgs = []\n",
    "    labels = []\n",
    "\n",
    "    all_img_paths = glob.glob(os.path.join(root_dir, '*/*.ppm'))\n",
    "    np.random.shuffle(all_img_paths)\n",
    "    for img_path in all_img_paths:\n",
    "        try:\n",
    "            img = preprocess_img(io.imread(img_path))\n",
    "            label = get_class(img_path)\n",
    "            imgs.append(img)\n",
    "            labels.append(label)\n",
    "\n",
    "            if len(imgs)%1000 == 0: print(\"Processed {}/{}\".format(len(imgs), len(all_img_paths)))\n",
    "        except (IOError, OSError):\n",
    "            print('missed', img_path)\n",
    "            pass\n",
    "\n",
    "    X = np.array(imgs, dtype='float32')\n",
    "    Y = np.eye(NUM_CLASSES, dtype='uint8')[labels]\n",
    "\n",
    "    with h5py.File('X.h5','w') as hf:\n",
    "        hf.create_dataset('imgs', data=X)\n",
    "        hf.create_dataset('labels', data=Y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Define Keras model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def cnn_model():\n",
    "    model = Sequential()\n",
    "\n",
    "    model.add(Conv2D(32, (3, 3), padding='same',\n",
    "                     input_shape=(3, IMG_SIZE, IMG_SIZE),\n",
    "                     activation='relu'))\n",
    "    model.add(Conv2D(32, (3, 3), activation='relu'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.2))\n",
    "\n",
    "    model.add(Conv2D(64, (3, 3), padding='same',\n",
    "                     activation='relu'))\n",
    "    model.add(Conv2D(64, (3, 3), activation='relu'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.2))\n",
    "\n",
    "    model.add(Conv2D(128, (3, 3), padding='same',\n",
    "                     activation='relu'))\n",
    "    model.add(Conv2D(128, (3, 3), activation='relu'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.2))\n",
    "\n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(512, activation='relu'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(NUM_CLASSES, activation='softmax'))\n",
    "    return model\n",
    "\n",
    "model = cnn_model()\n",
    "# let's train the model using SGD + momentum (how original).\n",
    "lr = 0.01\n",
    "sgd = SGD(lr=lr, decay=1e-6, momentum=0.9, nesterov=True)\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "          optimizer=sgd,\n",
    "          metrics=['accuracy'])\n",
    "\n",
    "\n",
    "def lr_schedule(epoch):\n",
    "    return lr*(0.1**int(epoch/10))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Start Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 31367 samples, validate on 7842 samples\n",
      "Epoch 1/30\n",
      "31367/31367 [==============================] - 30s - loss: 1.1502 - acc: 0.6723 - val_loss: 0.1262 - val_acc: 0.9616\n",
      "Epoch 2/30\n",
      "31367/31367 [==============================] - 32s - loss: 0.2143 - acc: 0.9359 - val_loss: 0.0653 - val_acc: 0.9809\n",
      "Epoch 3/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.1342 - acc: 0.9604 - val_loss: 0.0590 - val_acc: 0.9825\n",
      "Epoch 4/30\n",
      "31367/31367 [==============================] - 35s - loss: 0.0967 - acc: 0.9715 - val_loss: 0.0778 - val_acc: 0.9756\n",
      "Epoch 5/30\n",
      "31367/31367 [==============================] - 35s - loss: 0.0832 - acc: 0.9755 - val_loss: 0.0358 - val_acc: 0.9887\n",
      "Epoch 6/30\n",
      "31367/31367 [==============================] - 22s - loss: 0.0695 - acc: 0.9783 - val_loss: 0.0367 - val_acc: 0.9894\n",
      "Epoch 7/30\n",
      "31367/31367 [==============================] - 23s - loss: 0.0636 - acc: 0.9800 - val_loss: 0.0323 - val_acc: 0.9922\n",
      "Epoch 8/30\n",
      "31367/31367 [==============================] - 23s - loss: 0.0559 - acc: 0.9830 - val_loss: 0.0247 - val_acc: 0.9929\n",
      "Epoch 9/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0496 - acc: 0.9850 - val_loss: 0.0241 - val_acc: 0.9938\n",
      "Epoch 10/30\n",
      "31367/31367 [==============================] - 28s - loss: 0.0436 - acc: 0.9870 - val_loss: 0.0319 - val_acc: 0.9916\n",
      "Epoch 11/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0249 - acc: 0.9923 - val_loss: 0.0197 - val_acc: 0.9957\n",
      "Epoch 12/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0116 - acc: 0.9961 - val_loss: 0.0193 - val_acc: 0.9958\n",
      "Epoch 13/30\n",
      "31367/31367 [==============================] - 32s - loss: 0.0121 - acc: 0.9965 - val_loss: 0.0183 - val_acc: 0.9964\n",
      "Epoch 14/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0103 - acc: 0.9965 - val_loss: 0.0182 - val_acc: 0.9960\n",
      "Epoch 15/30\n",
      "31367/31367 [==============================] - 28s - loss: 0.0088 - acc: 0.9974 - val_loss: 0.0174 - val_acc: 0.9963\n",
      "Epoch 16/30\n",
      "31367/31367 [==============================] - 27s - loss: 0.0081 - acc: 0.9974 - val_loss: 0.0170 - val_acc: 0.9966\n",
      "Epoch 17/30\n",
      "31367/31367 [==============================] - 27s - loss: 0.0071 - acc: 0.9977 - val_loss: 0.0159 - val_acc: 0.9971\n",
      "Epoch 18/30\n",
      "31367/31367 [==============================] - 23s - loss: 0.0094 - acc: 0.9970 - val_loss: 0.0157 - val_acc: 0.9969\n",
      "Epoch 19/30\n",
      "31367/31367 [==============================] - 21s - loss: 0.0061 - acc: 0.9981 - val_loss: 0.0161 - val_acc: 0.9971\n",
      "Epoch 20/30\n",
      "31367/31367 [==============================] - 25s - loss: 0.0062 - acc: 0.9979 - val_loss: 0.0166 - val_acc: 0.9966\n",
      "Epoch 21/30\n",
      "31367/31367 [==============================] - 26s - loss: 0.0067 - acc: 0.9979 - val_loss: 0.0164 - val_acc: 0.9967\n",
      "Epoch 22/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0060 - acc: 0.9979 - val_loss: 0.0163 - val_acc: 0.9966\n",
      "Epoch 23/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0045 - acc: 0.9985 - val_loss: 0.0164 - val_acc: 0.9967\n",
      "Epoch 24/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0050 - acc: 0.9983 - val_loss: 0.0166 - val_acc: 0.9967\n",
      "Epoch 25/30\n",
      "31367/31367 [==============================] - 27s - loss: 0.0059 - acc: 0.9980 - val_loss: 0.0165 - val_acc: 0.9968\n",
      "Epoch 26/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0061 - acc: 0.9980 - val_loss: 0.0164 - val_acc: 0.9969\n",
      "Epoch 27/30\n",
      "31367/31367 [==============================] - 27s - loss: 0.0053 - acc: 0.9982 - val_loss: 0.0165 - val_acc: 0.9968\n",
      "Epoch 28/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0053 - acc: 0.9982 - val_loss: 0.0163 - val_acc: 0.9969\n",
      "Epoch 29/30\n",
      "31367/31367 [==============================] - 24s - loss: 0.0052 - acc: 0.9984 - val_loss: 0.0163 - val_acc: 0.9968\n",
      "Epoch 30/30\n",
      "31367/31367 [==============================] - 22s - loss: 0.0059 - acc: 0.9982 - val_loss: 0.0163 - val_acc: 0.9968\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7feaf3055898>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch_size = 32\n",
    "nb_epoch = 30\n",
    "\n",
    "model.fit(X, Y,\n",
    "          batch_size=batch_size,\n",
    "          epochs=nb_epoch,\n",
    "          validation_split=0.2,\n",
    "          shuffle=True,\n",
    "          callbacks=[LearningRateScheduler(lr_schedule),\n",
    "                    ModelCheckpoint('model.h5',save_best_only=True)]\n",
    "            )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Load Test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "test = pd.read_csv('GT-final_test.csv',sep=';')\n",
    "\n",
    "X_test = []\n",
    "y_test = []\n",
    "i = 0\n",
    "for file_name, class_id  in zip(list(test['Filename']), list(test['ClassId'])):\n",
    "    img_path = os.path.join('GTSRB/Final_Test/Images/',file_name)\n",
    "    X_test.append(preprocess_img(io.imread(img_path)))\n",
    "    y_test.append(class_id)\n",
    "    \n",
    "X_test = np.array(X_test)\n",
    "y_test = np.array(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12630/12630 [==============================] - 2s     \n",
      "Test accuracy = 0.9792557403008709\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict_classes(X_test)\n",
    "acc = np.sum(y_pred==y_test)/np.size(y_pred)\n",
    "print(\"Test accuracy = {}\".format(acc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# With Data augmentation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "from sklearn.cross_validation import train_test_split\n",
    "\n",
    "X_train, X_val, Y_train, Y_val = train_test_split(X, Y, test_size=0.2, random_state=42)\n",
    "\n",
    "datagen = ImageDataGenerator(featurewise_center=False, \n",
    "                            featurewise_std_normalization=False, \n",
    "                            width_shift_range=0.1,\n",
    "                            height_shift_range=0.1,\n",
    "                            zoom_range=0.2,\n",
    "                            shear_range=0.1,\n",
    "                            rotation_range=10.,)\n",
    "\n",
    "datagen.fit(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# Reinstallise models \n",
    "\n",
    "model = cnn_model()\n",
    "# let's train the model using SGD + momentum (how original).\n",
    "lr = 0.01\n",
    "sgd = SGD(lr=lr, decay=1e-6, momentum=0.9, nesterov=True)\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "          optimizer=sgd,\n",
    "          metrics=['accuracy'])\n",
    "\n",
    "\n",
    "def lr_schedule(epoch):\n",
    "    return lr*(0.1**int(epoch/10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "31367/31367 [==============================] - 27s - loss: 2.0311 - acc: 0.4023 - val_loss: 0.4975 - val_acc: 0.8258\n",
      "Epoch 2/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.7059 - acc: 0.7762 - val_loss: 0.2078 - val_acc: 0.9341\n",
      "Epoch 3/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.4047 - acc: 0.8763 - val_loss: 0.1255 - val_acc: 0.9621\n",
      "Epoch 4/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.2769 - acc: 0.9125 - val_loss: 0.0875 - val_acc: 0.9717\n",
      "Epoch 5/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.2266 - acc: 0.9308 - val_loss: 0.0459 - val_acc: 0.9865\n",
      "Epoch 6/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.1931 - acc: 0.9416 - val_loss: 0.0488 - val_acc: 0.9850\n",
      "Epoch 7/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.1737 - acc: 0.9477 - val_loss: 0.0383 - val_acc: 0.9881\n",
      "Epoch 8/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.1461 - acc: 0.9563 - val_loss: 0.0249 - val_acc: 0.9929\n",
      "Epoch 9/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.1406 - acc: 0.9591 - val_loss: 0.0257 - val_acc: 0.9926\n",
      "Epoch 10/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.1316 - acc: 0.9608 - val_loss: 0.0164 - val_acc: 0.9953\n",
      "Epoch 11/30\n",
      "31367/31367 [==============================] - 28s - loss: 0.0656 - acc: 0.9802 - val_loss: 0.0092 - val_acc: 0.9963\n",
      "Epoch 12/30\n",
      "31367/31367 [==============================] - 25s - loss: 0.0526 - acc: 0.9839 - val_loss: 0.0077 - val_acc: 0.9972\n",
      "Epoch 13/30\n",
      "31367/31367 [==============================] - 28s - loss: 0.0482 - acc: 0.9858 - val_loss: 0.0078 - val_acc: 0.9972\n",
      "Epoch 14/30\n",
      "31367/31367 [==============================] - 32s - loss: 0.0468 - acc: 0.9855 - val_loss: 0.0066 - val_acc: 0.9983\n",
      "Epoch 15/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.0446 - acc: 0.9864 - val_loss: 0.0059 - val_acc: 0.9982\n",
      "Epoch 16/30\n",
      "31367/31367 [==============================] - 32s - loss: 0.0394 - acc: 0.9879 - val_loss: 0.0058 - val_acc: 0.9985\n",
      "Epoch 17/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0414 - acc: 0.9877 - val_loss: 0.0061 - val_acc: 0.9981\n",
      "Epoch 18/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.0380 - acc: 0.9883 - val_loss: 0.0052 - val_acc: 0.9983\n",
      "Epoch 19/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.0354 - acc: 0.9894 - val_loss: 0.0055 - val_acc: 0.9981\n",
      "Epoch 20/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.0355 - acc: 0.9895 - val_loss: 0.0055 - val_acc: 0.9977\n",
      "Epoch 21/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.0357 - acc: 0.9895 - val_loss: 0.0050 - val_acc: 0.9981\n",
      "Epoch 22/30\n",
      "31367/31367 [==============================] - 32s - loss: 0.0329 - acc: 0.9898 - val_loss: 0.0048 - val_acc: 0.9983\n",
      "Epoch 23/30\n",
      "31367/31367 [==============================] - 25s - loss: 0.0329 - acc: 0.9905 - val_loss: 0.0048 - val_acc: 0.9981\n",
      "Epoch 24/30\n",
      "31367/31367 [==============================] - 29s - loss: 0.0323 - acc: 0.9901 - val_loss: 0.0047 - val_acc: 0.9982\n",
      "Epoch 25/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0357 - acc: 0.9892 - val_loss: 0.0046 - val_acc: 0.9983\n",
      "Epoch 26/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0317 - acc: 0.9899 - val_loss: 0.0046 - val_acc: 0.9982\n",
      "Epoch 27/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0320 - acc: 0.9904 - val_loss: 0.0045 - val_acc: 0.9982\n",
      "Epoch 28/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0340 - acc: 0.9895 - val_loss: 0.0046 - val_acc: 0.9986\n",
      "Epoch 29/30\n",
      "31367/31367 [==============================] - 30s - loss: 0.0319 - acc: 0.9904 - val_loss: 0.0045 - val_acc: 0.9987\n",
      "Epoch 30/30\n",
      "31367/31367 [==============================] - 31s - loss: 0.0318 - acc: 0.9896 - val_loss: 0.0044 - val_acc: 0.9987\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7fea770a6630>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nb_epoch = 30\n",
    "model.fit_generator(datagen.flow(X_train, Y_train, batch_size=batch_size),\n",
    "                            steps_per_epoch=X_train.shape[0],\n",
    "                            epochs=nb_epoch,\n",
    "                            validation_data=(X_val, Y_val),\n",
    "                            callbacks=[LearningRateScheduler(lr_schedule),\n",
    "                                       ModelCheckpoint('model.h5',save_best_only=True)]\n",
    "                           )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12630/12630 [==============================] - 3s     \n",
      "Test accuracy = 0.9828978622327791\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict_classes(X_test)\n",
    "acc = np.sum(y_pred==y_test)/np.size(y_pred)\n",
    "print(\"Test accuracy = {}\".format(acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "____________________________________________________________________________________________________\n",
      "Layer (type)                       Output Shape        Param #     Connected to                     \n",
      "====================================================================================================\n",
      "convolution2d_7 (Convolution2D)    (None, 32, 48, 48)  896         convolution2d_input_2[0][0]      \n",
      "____________________________________________________________________________________________________\n",
      "convolution2d_8 (Convolution2D)    (None, 32, 46, 46)  9248        convolution2d_7[0][0]            \n",
      "____________________________________________________________________________________________________\n",
      "maxpooling2d_4 (MaxPooling2D)      (None, 32, 23, 23)  0           convolution2d_8[0][0]            \n",
      "____________________________________________________________________________________________________\n",
      "dropout_5 (Dropout)                (None, 32, 23, 23)  0           maxpooling2d_4[0][0]             \n",
      "____________________________________________________________________________________________________\n",
      "convolution2d_9 (Convolution2D)    (None, 64, 23, 23)  18496       dropout_5[0][0]                  \n",
      "____________________________________________________________________________________________________\n",
      "convolution2d_10 (Convolution2D)   (None, 64, 21, 21)  36928       convolution2d_9[0][0]            \n",
      "____________________________________________________________________________________________________\n",
      "maxpooling2d_5 (MaxPooling2D)      (None, 64, 10, 10)  0           convolution2d_10[0][0]           \n",
      "____________________________________________________________________________________________________\n",
      "dropout_6 (Dropout)                (None, 64, 10, 10)  0           maxpooling2d_5[0][0]             \n",
      "____________________________________________________________________________________________________\n",
      "convolution2d_11 (Convolution2D)   (None, 128, 10, 10) 73856       dropout_6[0][0]                  \n",
      "____________________________________________________________________________________________________\n",
      "convolution2d_12 (Convolution2D)   (None, 128, 8, 8)   147584      convolution2d_11[0][0]           \n",
      "____________________________________________________________________________________________________\n",
      "maxpooling2d_6 (MaxPooling2D)      (None, 128, 4, 4)   0           convolution2d_12[0][0]           \n",
      "____________________________________________________________________________________________________\n",
      "dropout_7 (Dropout)                (None, 128, 4, 4)   0           maxpooling2d_6[0][0]             \n",
      "____________________________________________________________________________________________________\n",
      "flatten_2 (Flatten)                (None, 2048)        0           dropout_7[0][0]                  \n",
      "____________________________________________________________________________________________________\n",
      "dense_3 (Dense)                    (None, 512)         1049088     flatten_2[0][0]                  \n",
      "____________________________________________________________________________________________________\n",
      "dropout_8 (Dropout)                (None, 512)         0           dense_3[0][0]                    \n",
      "____________________________________________________________________________________________________\n",
      "dense_4 (Dense)                    (None, 43)          22059       dropout_8[0][0]                  \n",
      "====================================================================================================\n",
      "Total params: 1358155\n",
      "____________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1358155"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.count_params()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}