SE-Resnet Implementation

seres.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SE-ResNet Implementation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from __future__ import print_function\n",
    "from __future__ import absolute_import\n",
    "from __future__ import division\n",
    "from keras.models import Model\n",
    "from keras.layers import *\n",
    "from keras.regularizers import l2\n",
    "from keras.utils import conv_utils\n",
    "from keras.utils.data_utils import get_file\n",
    "from keras.engine.topology import get_source_inputs\n",
    "from keras_applications.imagenet_utils import _obtain_input_shape\n",
    "from keras_applications.resnet50 import preprocess_input\n",
    "from keras_applications.imagenet_utils import decode_predictions\n",
    "from keras import backend as K\n",
    "from datetime import datetime\n",
    "from visualization import *\n",
    "import time\n",
    "from keras.optimizers import SGD,Adam\n",
    "from keras.callbacks import ModelCheckpoint, LearningRateScheduler, EarlyStopping, ReduceLROnPlateau\n",
    "import numpy as np \n",
    "import pandas as pd\n",
    "from keras.utils import to_categorical"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we create out squeeze excite block, this is the main contribution of the paper and adds a global average pooling and 2 dense layers after the normal resnet block"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#two parameters: input and reduction ratio\n",
    "def squeeze_excite_block(input, ratio=16):\n",
    "    filter_kernels = input._keras_shape[-1]\n",
    "    z_shape = (1, 1, filter_kernels)\n",
    "    z = GlobalAveragePooling2D()(input)\n",
    "    z = Reshape(z_shape)(z)\n",
    "    s = Dense(filter_kernels//ratio, activation='relu', use_bias=False)(z)\n",
    "    s = Dense(filter_kernels, activation='sigmoid', use_bias=False)(s)\n",
    "    x = multiply([input, s])\n",
    "    return x"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we create out final netwokr block which is basically a bottleneck resnet block followed by the squeeze and excite block"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def se_resnet_block_bottleneck(input,channels,_strides=(1, 1)):\n",
    "    chan_axis=-1\n",
    "    if(input._keras_shape[-1]!=channels or _strides!=(1,1)):\n",
    "        input = Conv2D(channels, (1, 1), padding='same', kernel_initializer='he_normal',\n",
    "                      use_bias=False, strides=_strides)(input)\n",
    "    \n",
    "    x = Conv2D(channels, (1, 1), padding='same', kernel_initializer='he_normal',\n",
    "                      use_bias=False, strides=_strides)(input)\n",
    "    x = BatchNormalization(axis=chan_axis)(x)\n",
    "    x = Activation('relu')(x)\n",
    "    \n",
    "    \n",
    "    x = Conv2D(channels, (3, 3), padding='same', kernel_initializer='he_normal',\n",
    "                      use_bias=False, strides=_strides)(x)\n",
    "    x = BatchNormalization(axis=chan_axis)(x)\n",
    "    x = Activation('relu')(x)\n",
    "    \n",
    "    \n",
    "    x = Conv2D(channels, (1, 1), padding='same', kernel_initializer='he_normal',\n",
    "                      use_bias=False, strides=_strides)(x)\n",
    "    x = BatchNormalization(axis=chan_axis)(x)\n",
    "    x = Activation('relu')(x)\n",
    "    \n",
    "    \n",
    "    x = squeeze_excite_block(x)\n",
    "    out = add([x, input])\n",
    "    return out"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here we implement the architecture given in the paper diagram using for loops and adding blocks continuously"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def se_resnet(input,filters = [64,128,256,512],depth = [6,8,12,6],num_classes=10, weight_decay=1e-4):\n",
    "    chan_axis=-1\n",
    "    x = Conv2D(filters[0], (7, 7), padding='same', use_bias=False, strides=(2, 2),\n",
    "               kernel_initializer='he_normal', kernel_regularizer=l2(weight_decay))(input)\n",
    "    x = MaxPooling2D(pool_size=(2,2))(x)\n",
    "    for i in range(len(filters)):\n",
    "        x = se_resnet_block_bottleneck(x,filters[i],(2,2))\n",
    "        for j in range(depth[i]-1):\n",
    "            x = se_resnet_block_bottleneck(x,filters[i],(1,1))\n",
    "    x = GlobalAveragePooling2D()(x)\n",
    "    x = Dense(num_classes, activation='softmax', use_bias=False)(x)\n",
    "    return x"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the MNIST dataset here and create the create_model function that just uses the functions above to create our final model and return it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import mnist\n",
    "(train_x, train_y) , (test_x, test_y) = mnist.load_data()\n",
    "train_x = np.expand_dims(train_x,axis=3)\n",
    "#train_y = np.expand_dims(train_y,axis=1)\n",
    "train_y = to_categorical(train_y)\n",
    "test_x = np.expand_dims(test_x,axis=3)\n",
    "test_y = to_categorical(test_y)\n",
    "\n",
    "\n",
    "def create_model(input_shape = (28, 28, 1),filters = [64,128,256,512],depth = [6,8,12,6],num_classes=10, weight_decay=1e-4):\n",
    "    input = Input(shape = input_shape)\n",
    "    x = se_resnet(input,filters,depth,num_classes)\n",
    "    model = Model(input, x)\n",
    "    print(model.summary())\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28, 1)\n"
     ]
    }
   ],
   "source": [
    "print(train_x.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creating the model and printing the architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 28, 28, 1)    0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 14, 14, 64)   3136        input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)  (None, 7, 7, 64)     0           conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 4, 4, 64)     4096        max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_3 (Conv2D)               (None, 2, 2, 64)     4096        conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_1 (BatchNor (None, 2, 2, 64)     256         conv2d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_1 (Activation)       (None, 2, 2, 64)     0           batch_normalization_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_4 (Conv2D)               (None, 1, 1, 64)     36864       activation_1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_2 (BatchNor (None, 1, 1, 64)     256         conv2d_4[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_2 (Activation)       (None, 1, 1, 64)     0           batch_normalization_2[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_5 (Conv2D)               (None, 1, 1, 64)     4096        activation_2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_3 (BatchNor (None, 1, 1, 64)     256         conv2d_5[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_3 (Activation)       (None, 1, 1, 64)     0           batch_normalization_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_1 (Glo (None, 64)           0           activation_3[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "reshape_1 (Reshape)             (None, 1, 1, 64)     0           global_average_pooling2d_1[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 1, 1, 4)      256         reshape_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 1, 1, 64)     256         dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "multiply_1 (Multiply)           (None, 1, 1, 64)     0           activation_3[0][0]               \n",
      "                                                                 dense_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "add_1 (Add)                     (None, 4, 4, 64)     0           multiply_1[0][0]                 \n",
      "                                                                 conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_6 (Conv2D)               (None, 4, 4, 64)     4096        add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_4 (BatchNor (None, 4, 4, 64)     256         conv2d_6[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_4 (Activation)       (None, 4, 4, 64)     0           batch_normalization_4[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_7 (Conv2D)               (None, 4, 4, 64)     36864       activation_4[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_5 (BatchNor (None, 4, 4, 64)     256         conv2d_7[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_5 (Activation)       (None, 4, 4, 64)     0           batch_normalization_5[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_8 (Conv2D)               (None, 4, 4, 64)     4096        activation_5[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_6 (BatchNor (None, 4, 4, 64)     256         conv2d_8[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_6 (Activation)       (None, 4, 4, 64)     0           batch_normalization_6[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_2 (Glo (None, 64)           0           activation_6[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "reshape_2 (Reshape)             (None, 1, 1, 64)     0           global_average_pooling2d_2[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_3 (Dense)                 (None, 1, 1, 4)      256         reshape_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_4 (Dense)                 (None, 1, 1, 64)     256         dense_3[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "multiply_2 (Multiply)           (None, 4, 4, 64)     0           activation_6[0][0]               \n",
      "                                                                 dense_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "add_2 (Add)                     (None, 4, 4, 64)     0           multiply_2[0][0]                 \n",
      "                                                                 add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_9 (Conv2D)               (None, 4, 4, 64)     4096        add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_7 (BatchNor (None, 4, 4, 64)     256         conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_7 (Activation)       (None, 4, 4, 64)     0           batch_normalization_7[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_10 (Conv2D)              (None, 4, 4, 64)     36864       activation_7[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_8 (BatchNor (None, 4, 4, 64)     256         conv2d_10[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_8 (Activation)       (None, 4, 4, 64)     0           batch_normalization_8[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_11 (Conv2D)              (None, 4, 4, 64)     4096        activation_8[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_9 (BatchNor (None, 4, 4, 64)     256         conv2d_11[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_9 (Activation)       (None, 4, 4, 64)     0           batch_normalization_9[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_3 (Glo (None, 64)           0           activation_9[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "reshape_3 (Reshape)             (None, 1, 1, 64)     0           global_average_pooling2d_3[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_5 (Dense)                 (None, 1, 1, 4)      256         reshape_3[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_6 (Dense)                 (None, 1, 1, 64)     256         dense_5[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "multiply_3 (Multiply)           (None, 4, 4, 64)     0           activation_9[0][0]               \n",
      "                                                                 dense_6[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "add_3 (Add)                     (None, 4, 4, 64)     0           multiply_3[0][0]                 \n",
      "                                                                 add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_12 (Conv2D)              (None, 4, 4, 64)     4096        add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_10 (BatchNo (None, 4, 4, 64)     256         conv2d_12[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_10 (Activation)      (None, 4, 4, 64)     0           batch_normalization_10[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_13 (Conv2D)              (None, 4, 4, 64)     36864       activation_10[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_11 (BatchNo (None, 4, 4, 64)     256         conv2d_13[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_11 (Activation)      (None, 4, 4, 64)     0           batch_normalization_11[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_14 (Conv2D)              (None, 4, 4, 64)     4096        activation_11[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_12 (BatchNo (None, 4, 4, 64)     256         conv2d_14[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_12 (Activation)      (None, 4, 4, 64)     0           batch_normalization_12[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_4 (Glo (None, 64)           0           activation_12[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_4 (Reshape)             (None, 1, 1, 64)     0           global_average_pooling2d_4[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_7 (Dense)                 (None, 1, 1, 4)      256         reshape_4[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_8 (Dense)                 (None, 1, 1, 64)     256         dense_7[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "multiply_4 (Multiply)           (None, 4, 4, 64)     0           activation_12[0][0]              \n",
      "                                                                 dense_8[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "add_4 (Add)                     (None, 4, 4, 64)     0           multiply_4[0][0]                 \n",
      "                                                                 add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_15 (Conv2D)              (None, 4, 4, 64)     4096        add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_13 (BatchNo (None, 4, 4, 64)     256         conv2d_15[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_13 (Activation)      (None, 4, 4, 64)     0           batch_normalization_13[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_16 (Conv2D)              (None, 4, 4, 64)     36864       activation_13[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_14 (BatchNo (None, 4, 4, 64)     256         conv2d_16[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_14 (Activation)      (None, 4, 4, 64)     0           batch_normalization_14[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_17 (Conv2D)              (None, 4, 4, 64)     4096        activation_14[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_15 (BatchNo (None, 4, 4, 64)     256         conv2d_17[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_15 (Activation)      (None, 4, 4, 64)     0           batch_normalization_15[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_5 (Glo (None, 64)           0           activation_15[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_5 (Reshape)             (None, 1, 1, 64)     0           global_average_pooling2d_5[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_9 (Dense)                 (None, 1, 1, 4)      256         reshape_5[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_10 (Dense)                (None, 1, 1, 64)     256         dense_9[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "multiply_5 (Multiply)           (None, 4, 4, 64)     0           activation_15[0][0]              \n",
      "                                                                 dense_10[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_5 (Add)                     (None, 4, 4, 64)     0           multiply_5[0][0]                 \n",
      "                                                                 add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_18 (Conv2D)              (None, 4, 4, 64)     4096        add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_16 (BatchNo (None, 4, 4, 64)     256         conv2d_18[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_16 (Activation)      (None, 4, 4, 64)     0           batch_normalization_16[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_19 (Conv2D)              (None, 4, 4, 64)     36864       activation_16[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_17 (BatchNo (None, 4, 4, 64)     256         conv2d_19[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_17 (Activation)      (None, 4, 4, 64)     0           batch_normalization_17[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_20 (Conv2D)              (None, 4, 4, 64)     4096        activation_17[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_18 (BatchNo (None, 4, 4, 64)     256         conv2d_20[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_18 (Activation)      (None, 4, 4, 64)     0           batch_normalization_18[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_6 (Glo (None, 64)           0           activation_18[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_6 (Reshape)             (None, 1, 1, 64)     0           global_average_pooling2d_6[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_11 (Dense)                (None, 1, 1, 4)      256         reshape_6[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_12 (Dense)                (None, 1, 1, 64)     256         dense_11[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_6 (Multiply)           (None, 4, 4, 64)     0           activation_18[0][0]              \n",
      "                                                                 dense_12[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_6 (Add)                     (None, 4, 4, 64)     0           multiply_6[0][0]                 \n",
      "                                                                 add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_21 (Conv2D)              (None, 2, 2, 128)    8192        add_6[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_22 (Conv2D)              (None, 1, 1, 128)    16384       conv2d_21[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_19 (BatchNo (None, 1, 1, 128)    512         conv2d_22[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_19 (Activation)      (None, 1, 1, 128)    0           batch_normalization_19[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_23 (Conv2D)              (None, 1, 1, 128)    147456      activation_19[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_20 (BatchNo (None, 1, 1, 128)    512         conv2d_23[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_20 (Activation)      (None, 1, 1, 128)    0           batch_normalization_20[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_24 (Conv2D)              (None, 1, 1, 128)    16384       activation_20[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_21 (BatchNo (None, 1, 1, 128)    512         conv2d_24[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_21 (Activation)      (None, 1, 1, 128)    0           batch_normalization_21[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_7 (Glo (None, 128)          0           activation_21[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_7 (Reshape)             (None, 1, 1, 128)    0           global_average_pooling2d_7[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_13 (Dense)                (None, 1, 1, 8)      1024        reshape_7[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_14 (Dense)                (None, 1, 1, 128)    1024        dense_13[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_7 (Multiply)           (None, 1, 1, 128)    0           activation_21[0][0]              \n",
      "                                                                 dense_14[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_7 (Add)                     (None, 2, 2, 128)    0           multiply_7[0][0]                 \n",
      "                                                                 conv2d_21[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_25 (Conv2D)              (None, 2, 2, 128)    16384       add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_22 (BatchNo (None, 2, 2, 128)    512         conv2d_25[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_22 (Activation)      (None, 2, 2, 128)    0           batch_normalization_22[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_26 (Conv2D)              (None, 2, 2, 128)    147456      activation_22[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_23 (BatchNo (None, 2, 2, 128)    512         conv2d_26[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_23 (Activation)      (None, 2, 2, 128)    0           batch_normalization_23[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_27 (Conv2D)              (None, 2, 2, 128)    16384       activation_23[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_24 (BatchNo (None, 2, 2, 128)    512         conv2d_27[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_24 (Activation)      (None, 2, 2, 128)    0           batch_normalization_24[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_8 (Glo (None, 128)          0           activation_24[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_8 (Reshape)             (None, 1, 1, 128)    0           global_average_pooling2d_8[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_15 (Dense)                (None, 1, 1, 8)      1024        reshape_8[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_16 (Dense)                (None, 1, 1, 128)    1024        dense_15[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_8 (Multiply)           (None, 2, 2, 128)    0           activation_24[0][0]              \n",
      "                                                                 dense_16[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_8 (Add)                     (None, 2, 2, 128)    0           multiply_8[0][0]                 \n",
      "                                                                 add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_28 (Conv2D)              (None, 2, 2, 128)    16384       add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_25 (BatchNo (None, 2, 2, 128)    512         conv2d_28[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_25 (Activation)      (None, 2, 2, 128)    0           batch_normalization_25[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_29 (Conv2D)              (None, 2, 2, 128)    147456      activation_25[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_26 (BatchNo (None, 2, 2, 128)    512         conv2d_29[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_26 (Activation)      (None, 2, 2, 128)    0           batch_normalization_26[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_30 (Conv2D)              (None, 2, 2, 128)    16384       activation_26[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_27 (BatchNo (None, 2, 2, 128)    512         conv2d_30[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_27 (Activation)      (None, 2, 2, 128)    0           batch_normalization_27[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_9 (Glo (None, 128)          0           activation_27[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_9 (Reshape)             (None, 1, 1, 128)    0           global_average_pooling2d_9[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "dense_17 (Dense)                (None, 1, 1, 8)      1024        reshape_9[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_18 (Dense)                (None, 1, 1, 128)    1024        dense_17[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_9 (Multiply)           (None, 2, 2, 128)    0           activation_27[0][0]              \n",
      "                                                                 dense_18[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_9 (Add)                     (None, 2, 2, 128)    0           multiply_9[0][0]                 \n",
      "                                                                 add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_31 (Conv2D)              (None, 2, 2, 128)    16384       add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_28 (BatchNo (None, 2, 2, 128)    512         conv2d_31[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_28 (Activation)      (None, 2, 2, 128)    0           batch_normalization_28[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_32 (Conv2D)              (None, 2, 2, 128)    147456      activation_28[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_29 (BatchNo (None, 2, 2, 128)    512         conv2d_32[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_29 (Activation)      (None, 2, 2, 128)    0           batch_normalization_29[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_33 (Conv2D)              (None, 2, 2, 128)    16384       activation_29[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_30 (BatchNo (None, 2, 2, 128)    512         conv2d_33[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_30 (Activation)      (None, 2, 2, 128)    0           batch_normalization_30[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_10 (Gl (None, 128)          0           activation_30[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_10 (Reshape)            (None, 1, 1, 128)    0           global_average_pooling2d_10[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_19 (Dense)                (None, 1, 1, 8)      1024        reshape_10[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_20 (Dense)                (None, 1, 1, 128)    1024        dense_19[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_10 (Multiply)          (None, 2, 2, 128)    0           activation_30[0][0]              \n",
      "                                                                 dense_20[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_10 (Add)                    (None, 2, 2, 128)    0           multiply_10[0][0]                \n",
      "                                                                 add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_34 (Conv2D)              (None, 2, 2, 128)    16384       add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_31 (BatchNo (None, 2, 2, 128)    512         conv2d_34[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_31 (Activation)      (None, 2, 2, 128)    0           batch_normalization_31[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_35 (Conv2D)              (None, 2, 2, 128)    147456      activation_31[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_32 (BatchNo (None, 2, 2, 128)    512         conv2d_35[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_32 (Activation)      (None, 2, 2, 128)    0           batch_normalization_32[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_36 (Conv2D)              (None, 2, 2, 128)    16384       activation_32[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_33 (BatchNo (None, 2, 2, 128)    512         conv2d_36[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_33 (Activation)      (None, 2, 2, 128)    0           batch_normalization_33[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_11 (Gl (None, 128)          0           activation_33[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_11 (Reshape)            (None, 1, 1, 128)    0           global_average_pooling2d_11[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_21 (Dense)                (None, 1, 1, 8)      1024        reshape_11[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_22 (Dense)                (None, 1, 1, 128)    1024        dense_21[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_11 (Multiply)          (None, 2, 2, 128)    0           activation_33[0][0]              \n",
      "                                                                 dense_22[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_11 (Add)                    (None, 2, 2, 128)    0           multiply_11[0][0]                \n",
      "                                                                 add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_37 (Conv2D)              (None, 2, 2, 128)    16384       add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_34 (BatchNo (None, 2, 2, 128)    512         conv2d_37[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_34 (Activation)      (None, 2, 2, 128)    0           batch_normalization_34[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_38 (Conv2D)              (None, 2, 2, 128)    147456      activation_34[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_35 (BatchNo (None, 2, 2, 128)    512         conv2d_38[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_35 (Activation)      (None, 2, 2, 128)    0           batch_normalization_35[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_39 (Conv2D)              (None, 2, 2, 128)    16384       activation_35[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_36 (BatchNo (None, 2, 2, 128)    512         conv2d_39[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_36 (Activation)      (None, 2, 2, 128)    0           batch_normalization_36[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_12 (Gl (None, 128)          0           activation_36[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_12 (Reshape)            (None, 1, 1, 128)    0           global_average_pooling2d_12[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_23 (Dense)                (None, 1, 1, 8)      1024        reshape_12[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_24 (Dense)                (None, 1, 1, 128)    1024        dense_23[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_12 (Multiply)          (None, 2, 2, 128)    0           activation_36[0][0]              \n",
      "                                                                 dense_24[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_12 (Add)                    (None, 2, 2, 128)    0           multiply_12[0][0]                \n",
      "                                                                 add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_40 (Conv2D)              (None, 2, 2, 128)    16384       add_12[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_37 (BatchNo (None, 2, 2, 128)    512         conv2d_40[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_37 (Activation)      (None, 2, 2, 128)    0           batch_normalization_37[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_41 (Conv2D)              (None, 2, 2, 128)    147456      activation_37[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_38 (BatchNo (None, 2, 2, 128)    512         conv2d_41[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_38 (Activation)      (None, 2, 2, 128)    0           batch_normalization_38[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_42 (Conv2D)              (None, 2, 2, 128)    16384       activation_38[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_39 (BatchNo (None, 2, 2, 128)    512         conv2d_42[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_39 (Activation)      (None, 2, 2, 128)    0           batch_normalization_39[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_13 (Gl (None, 128)          0           activation_39[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_13 (Reshape)            (None, 1, 1, 128)    0           global_average_pooling2d_13[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_25 (Dense)                (None, 1, 1, 8)      1024        reshape_13[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_26 (Dense)                (None, 1, 1, 128)    1024        dense_25[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_13 (Multiply)          (None, 2, 2, 128)    0           activation_39[0][0]              \n",
      "                                                                 dense_26[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_13 (Add)                    (None, 2, 2, 128)    0           multiply_13[0][0]                \n",
      "                                                                 add_12[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_43 (Conv2D)              (None, 2, 2, 128)    16384       add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_40 (BatchNo (None, 2, 2, 128)    512         conv2d_43[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_40 (Activation)      (None, 2, 2, 128)    0           batch_normalization_40[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_44 (Conv2D)              (None, 2, 2, 128)    147456      activation_40[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_41 (BatchNo (None, 2, 2, 128)    512         conv2d_44[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_41 (Activation)      (None, 2, 2, 128)    0           batch_normalization_41[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_45 (Conv2D)              (None, 2, 2, 128)    16384       activation_41[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_42 (BatchNo (None, 2, 2, 128)    512         conv2d_45[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_42 (Activation)      (None, 2, 2, 128)    0           batch_normalization_42[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_14 (Gl (None, 128)          0           activation_42[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_14 (Reshape)            (None, 1, 1, 128)    0           global_average_pooling2d_14[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_27 (Dense)                (None, 1, 1, 8)      1024        reshape_14[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_28 (Dense)                (None, 1, 1, 128)    1024        dense_27[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_14 (Multiply)          (None, 2, 2, 128)    0           activation_42[0][0]              \n",
      "                                                                 dense_28[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_14 (Add)                    (None, 2, 2, 128)    0           multiply_14[0][0]                \n",
      "                                                                 add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_46 (Conv2D)              (None, 1, 1, 256)    32768       add_14[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_47 (Conv2D)              (None, 1, 1, 256)    65536       conv2d_46[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_43 (BatchNo (None, 1, 1, 256)    1024        conv2d_47[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_43 (Activation)      (None, 1, 1, 256)    0           batch_normalization_43[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_48 (Conv2D)              (None, 1, 1, 256)    589824      activation_43[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_44 (BatchNo (None, 1, 1, 256)    1024        conv2d_48[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_44 (Activation)      (None, 1, 1, 256)    0           batch_normalization_44[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_49 (Conv2D)              (None, 1, 1, 256)    65536       activation_44[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_45 (BatchNo (None, 1, 1, 256)    1024        conv2d_49[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_45 (Activation)      (None, 1, 1, 256)    0           batch_normalization_45[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_15 (Gl (None, 256)          0           activation_45[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_15 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_15[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_29 (Dense)                (None, 1, 1, 16)     4096        reshape_15[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_30 (Dense)                (None, 1, 1, 256)    4096        dense_29[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_15 (Multiply)          (None, 1, 1, 256)    0           activation_45[0][0]              \n",
      "                                                                 dense_30[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_15 (Add)                    (None, 1, 1, 256)    0           multiply_15[0][0]                \n",
      "                                                                 conv2d_46[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_50 (Conv2D)              (None, 1, 1, 256)    65536       add_15[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_46 (BatchNo (None, 1, 1, 256)    1024        conv2d_50[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_46 (Activation)      (None, 1, 1, 256)    0           batch_normalization_46[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_51 (Conv2D)              (None, 1, 1, 256)    589824      activation_46[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_47 (BatchNo (None, 1, 1, 256)    1024        conv2d_51[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_47 (Activation)      (None, 1, 1, 256)    0           batch_normalization_47[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_52 (Conv2D)              (None, 1, 1, 256)    65536       activation_47[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_48 (BatchNo (None, 1, 1, 256)    1024        conv2d_52[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_48 (Activation)      (None, 1, 1, 256)    0           batch_normalization_48[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_16 (Gl (None, 256)          0           activation_48[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_16 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_16[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_31 (Dense)                (None, 1, 1, 16)     4096        reshape_16[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_32 (Dense)                (None, 1, 1, 256)    4096        dense_31[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_16 (Multiply)          (None, 1, 1, 256)    0           activation_48[0][0]              \n",
      "                                                                 dense_32[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_16 (Add)                    (None, 1, 1, 256)    0           multiply_16[0][0]                \n",
      "                                                                 add_15[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_53 (Conv2D)              (None, 1, 1, 256)    65536       add_16[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_49 (BatchNo (None, 1, 1, 256)    1024        conv2d_53[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_49 (Activation)      (None, 1, 1, 256)    0           batch_normalization_49[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_54 (Conv2D)              (None, 1, 1, 256)    589824      activation_49[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_50 (BatchNo (None, 1, 1, 256)    1024        conv2d_54[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_50 (Activation)      (None, 1, 1, 256)    0           batch_normalization_50[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_55 (Conv2D)              (None, 1, 1, 256)    65536       activation_50[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_51 (BatchNo (None, 1, 1, 256)    1024        conv2d_55[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_51 (Activation)      (None, 1, 1, 256)    0           batch_normalization_51[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_17 (Gl (None, 256)          0           activation_51[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_17 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_17[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_33 (Dense)                (None, 1, 1, 16)     4096        reshape_17[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_34 (Dense)                (None, 1, 1, 256)    4096        dense_33[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_17 (Multiply)          (None, 1, 1, 256)    0           activation_51[0][0]              \n",
      "                                                                 dense_34[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_17 (Add)                    (None, 1, 1, 256)    0           multiply_17[0][0]                \n",
      "                                                                 add_16[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_56 (Conv2D)              (None, 1, 1, 256)    65536       add_17[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_52 (BatchNo (None, 1, 1, 256)    1024        conv2d_56[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_52 (Activation)      (None, 1, 1, 256)    0           batch_normalization_52[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_57 (Conv2D)              (None, 1, 1, 256)    589824      activation_52[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_53 (BatchNo (None, 1, 1, 256)    1024        conv2d_57[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_53 (Activation)      (None, 1, 1, 256)    0           batch_normalization_53[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_58 (Conv2D)              (None, 1, 1, 256)    65536       activation_53[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_54 (BatchNo (None, 1, 1, 256)    1024        conv2d_58[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_54 (Activation)      (None, 1, 1, 256)    0           batch_normalization_54[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_18 (Gl (None, 256)          0           activation_54[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_18 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_18[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_35 (Dense)                (None, 1, 1, 16)     4096        reshape_18[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_36 (Dense)                (None, 1, 1, 256)    4096        dense_35[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_18 (Multiply)          (None, 1, 1, 256)    0           activation_54[0][0]              \n",
      "                                                                 dense_36[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_18 (Add)                    (None, 1, 1, 256)    0           multiply_18[0][0]                \n",
      "                                                                 add_17[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_59 (Conv2D)              (None, 1, 1, 256)    65536       add_18[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_55 (BatchNo (None, 1, 1, 256)    1024        conv2d_59[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_55 (Activation)      (None, 1, 1, 256)    0           batch_normalization_55[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_60 (Conv2D)              (None, 1, 1, 256)    589824      activation_55[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_56 (BatchNo (None, 1, 1, 256)    1024        conv2d_60[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_56 (Activation)      (None, 1, 1, 256)    0           batch_normalization_56[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_61 (Conv2D)              (None, 1, 1, 256)    65536       activation_56[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_57 (BatchNo (None, 1, 1, 256)    1024        conv2d_61[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_57 (Activation)      (None, 1, 1, 256)    0           batch_normalization_57[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_19 (Gl (None, 256)          0           activation_57[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_19 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_19[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_37 (Dense)                (None, 1, 1, 16)     4096        reshape_19[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_38 (Dense)                (None, 1, 1, 256)    4096        dense_37[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_19 (Multiply)          (None, 1, 1, 256)    0           activation_57[0][0]              \n",
      "                                                                 dense_38[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_19 (Add)                    (None, 1, 1, 256)    0           multiply_19[0][0]                \n",
      "                                                                 add_18[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_62 (Conv2D)              (None, 1, 1, 256)    65536       add_19[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_58 (BatchNo (None, 1, 1, 256)    1024        conv2d_62[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_58 (Activation)      (None, 1, 1, 256)    0           batch_normalization_58[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_63 (Conv2D)              (None, 1, 1, 256)    589824      activation_58[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_59 (BatchNo (None, 1, 1, 256)    1024        conv2d_63[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_59 (Activation)      (None, 1, 1, 256)    0           batch_normalization_59[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_64 (Conv2D)              (None, 1, 1, 256)    65536       activation_59[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_60 (BatchNo (None, 1, 1, 256)    1024        conv2d_64[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_60 (Activation)      (None, 1, 1, 256)    0           batch_normalization_60[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_20 (Gl (None, 256)          0           activation_60[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_20 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_20[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_39 (Dense)                (None, 1, 1, 16)     4096        reshape_20[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_40 (Dense)                (None, 1, 1, 256)    4096        dense_39[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_20 (Multiply)          (None, 1, 1, 256)    0           activation_60[0][0]              \n",
      "                                                                 dense_40[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_20 (Add)                    (None, 1, 1, 256)    0           multiply_20[0][0]                \n",
      "                                                                 add_19[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_65 (Conv2D)              (None, 1, 1, 256)    65536       add_20[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_61 (BatchNo (None, 1, 1, 256)    1024        conv2d_65[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_61 (Activation)      (None, 1, 1, 256)    0           batch_normalization_61[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_66 (Conv2D)              (None, 1, 1, 256)    589824      activation_61[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_62 (BatchNo (None, 1, 1, 256)    1024        conv2d_66[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_62 (Activation)      (None, 1, 1, 256)    0           batch_normalization_62[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_67 (Conv2D)              (None, 1, 1, 256)    65536       activation_62[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_63 (BatchNo (None, 1, 1, 256)    1024        conv2d_67[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_63 (Activation)      (None, 1, 1, 256)    0           batch_normalization_63[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_21 (Gl (None, 256)          0           activation_63[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_21 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_21[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_41 (Dense)                (None, 1, 1, 16)     4096        reshape_21[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_42 (Dense)                (None, 1, 1, 256)    4096        dense_41[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_21 (Multiply)          (None, 1, 1, 256)    0           activation_63[0][0]              \n",
      "                                                                 dense_42[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_21 (Add)                    (None, 1, 1, 256)    0           multiply_21[0][0]                \n",
      "                                                                 add_20[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_68 (Conv2D)              (None, 1, 1, 256)    65536       add_21[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_64 (BatchNo (None, 1, 1, 256)    1024        conv2d_68[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_64 (Activation)      (None, 1, 1, 256)    0           batch_normalization_64[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_69 (Conv2D)              (None, 1, 1, 256)    589824      activation_64[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_65 (BatchNo (None, 1, 1, 256)    1024        conv2d_69[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_65 (Activation)      (None, 1, 1, 256)    0           batch_normalization_65[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_70 (Conv2D)              (None, 1, 1, 256)    65536       activation_65[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_66 (BatchNo (None, 1, 1, 256)    1024        conv2d_70[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_66 (Activation)      (None, 1, 1, 256)    0           batch_normalization_66[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_22 (Gl (None, 256)          0           activation_66[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_22 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_22[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_43 (Dense)                (None, 1, 1, 16)     4096        reshape_22[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_44 (Dense)                (None, 1, 1, 256)    4096        dense_43[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_22 (Multiply)          (None, 1, 1, 256)    0           activation_66[0][0]              \n",
      "                                                                 dense_44[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_22 (Add)                    (None, 1, 1, 256)    0           multiply_22[0][0]                \n",
      "                                                                 add_21[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_71 (Conv2D)              (None, 1, 1, 256)    65536       add_22[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_67 (BatchNo (None, 1, 1, 256)    1024        conv2d_71[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_67 (Activation)      (None, 1, 1, 256)    0           batch_normalization_67[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_72 (Conv2D)              (None, 1, 1, 256)    589824      activation_67[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_68 (BatchNo (None, 1, 1, 256)    1024        conv2d_72[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_68 (Activation)      (None, 1, 1, 256)    0           batch_normalization_68[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_73 (Conv2D)              (None, 1, 1, 256)    65536       activation_68[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_69 (BatchNo (None, 1, 1, 256)    1024        conv2d_73[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_69 (Activation)      (None, 1, 1, 256)    0           batch_normalization_69[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_23 (Gl (None, 256)          0           activation_69[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_23 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_23[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_45 (Dense)                (None, 1, 1, 16)     4096        reshape_23[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_46 (Dense)                (None, 1, 1, 256)    4096        dense_45[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_23 (Multiply)          (None, 1, 1, 256)    0           activation_69[0][0]              \n",
      "                                                                 dense_46[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_23 (Add)                    (None, 1, 1, 256)    0           multiply_23[0][0]                \n",
      "                                                                 add_22[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_74 (Conv2D)              (None, 1, 1, 256)    65536       add_23[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_70 (BatchNo (None, 1, 1, 256)    1024        conv2d_74[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_70 (Activation)      (None, 1, 1, 256)    0           batch_normalization_70[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_75 (Conv2D)              (None, 1, 1, 256)    589824      activation_70[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_71 (BatchNo (None, 1, 1, 256)    1024        conv2d_75[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_71 (Activation)      (None, 1, 1, 256)    0           batch_normalization_71[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_76 (Conv2D)              (None, 1, 1, 256)    65536       activation_71[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_72 (BatchNo (None, 1, 1, 256)    1024        conv2d_76[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_72 (Activation)      (None, 1, 1, 256)    0           batch_normalization_72[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_24 (Gl (None, 256)          0           activation_72[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_24 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_24[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_47 (Dense)                (None, 1, 1, 16)     4096        reshape_24[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_48 (Dense)                (None, 1, 1, 256)    4096        dense_47[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_24 (Multiply)          (None, 1, 1, 256)    0           activation_72[0][0]              \n",
      "                                                                 dense_48[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_24 (Add)                    (None, 1, 1, 256)    0           multiply_24[0][0]                \n",
      "                                                                 add_23[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_77 (Conv2D)              (None, 1, 1, 256)    65536       add_24[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_73 (BatchNo (None, 1, 1, 256)    1024        conv2d_77[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_73 (Activation)      (None, 1, 1, 256)    0           batch_normalization_73[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_78 (Conv2D)              (None, 1, 1, 256)    589824      activation_73[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_74 (BatchNo (None, 1, 1, 256)    1024        conv2d_78[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_74 (Activation)      (None, 1, 1, 256)    0           batch_normalization_74[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_79 (Conv2D)              (None, 1, 1, 256)    65536       activation_74[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_75 (BatchNo (None, 1, 1, 256)    1024        conv2d_79[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_75 (Activation)      (None, 1, 1, 256)    0           batch_normalization_75[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_25 (Gl (None, 256)          0           activation_75[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_25 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_25[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_49 (Dense)                (None, 1, 1, 16)     4096        reshape_25[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_50 (Dense)                (None, 1, 1, 256)    4096        dense_49[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_25 (Multiply)          (None, 1, 1, 256)    0           activation_75[0][0]              \n",
      "                                                                 dense_50[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_25 (Add)                    (None, 1, 1, 256)    0           multiply_25[0][0]                \n",
      "                                                                 add_24[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_80 (Conv2D)              (None, 1, 1, 256)    65536       add_25[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_76 (BatchNo (None, 1, 1, 256)    1024        conv2d_80[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_76 (Activation)      (None, 1, 1, 256)    0           batch_normalization_76[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_81 (Conv2D)              (None, 1, 1, 256)    589824      activation_76[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_77 (BatchNo (None, 1, 1, 256)    1024        conv2d_81[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_77 (Activation)      (None, 1, 1, 256)    0           batch_normalization_77[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_82 (Conv2D)              (None, 1, 1, 256)    65536       activation_77[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_78 (BatchNo (None, 1, 1, 256)    1024        conv2d_82[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_78 (Activation)      (None, 1, 1, 256)    0           batch_normalization_78[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_26 (Gl (None, 256)          0           activation_78[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_26 (Reshape)            (None, 1, 1, 256)    0           global_average_pooling2d_26[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_51 (Dense)                (None, 1, 1, 16)     4096        reshape_26[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_52 (Dense)                (None, 1, 1, 256)    4096        dense_51[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_26 (Multiply)          (None, 1, 1, 256)    0           activation_78[0][0]              \n",
      "                                                                 dense_52[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_26 (Add)                    (None, 1, 1, 256)    0           multiply_26[0][0]                \n",
      "                                                                 add_25[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_83 (Conv2D)              (None, 1, 1, 512)    131072      add_26[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_84 (Conv2D)              (None, 1, 1, 512)    262144      conv2d_83[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_79 (BatchNo (None, 1, 1, 512)    2048        conv2d_84[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_79 (Activation)      (None, 1, 1, 512)    0           batch_normalization_79[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_85 (Conv2D)              (None, 1, 1, 512)    2359296     activation_79[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_80 (BatchNo (None, 1, 1, 512)    2048        conv2d_85[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_80 (Activation)      (None, 1, 1, 512)    0           batch_normalization_80[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_86 (Conv2D)              (None, 1, 1, 512)    262144      activation_80[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_81 (BatchNo (None, 1, 1, 512)    2048        conv2d_86[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_81 (Activation)      (None, 1, 1, 512)    0           batch_normalization_81[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_27 (Gl (None, 512)          0           activation_81[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_27 (Reshape)            (None, 1, 1, 512)    0           global_average_pooling2d_27[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_53 (Dense)                (None, 1, 1, 32)     16384       reshape_27[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_54 (Dense)                (None, 1, 1, 512)    16384       dense_53[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_27 (Multiply)          (None, 1, 1, 512)    0           activation_81[0][0]              \n",
      "                                                                 dense_54[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_27 (Add)                    (None, 1, 1, 512)    0           multiply_27[0][0]                \n",
      "                                                                 conv2d_83[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_87 (Conv2D)              (None, 1, 1, 512)    262144      add_27[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_82 (BatchNo (None, 1, 1, 512)    2048        conv2d_87[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_82 (Activation)      (None, 1, 1, 512)    0           batch_normalization_82[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_88 (Conv2D)              (None, 1, 1, 512)    2359296     activation_82[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_83 (BatchNo (None, 1, 1, 512)    2048        conv2d_88[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_83 (Activation)      (None, 1, 1, 512)    0           batch_normalization_83[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_89 (Conv2D)              (None, 1, 1, 512)    262144      activation_83[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_84 (BatchNo (None, 1, 1, 512)    2048        conv2d_89[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_84 (Activation)      (None, 1, 1, 512)    0           batch_normalization_84[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_28 (Gl (None, 512)          0           activation_84[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_28 (Reshape)            (None, 1, 1, 512)    0           global_average_pooling2d_28[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_55 (Dense)                (None, 1, 1, 32)     16384       reshape_28[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_56 (Dense)                (None, 1, 1, 512)    16384       dense_55[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_28 (Multiply)          (None, 1, 1, 512)    0           activation_84[0][0]              \n",
      "                                                                 dense_56[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_28 (Add)                    (None, 1, 1, 512)    0           multiply_28[0][0]                \n",
      "                                                                 add_27[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_90 (Conv2D)              (None, 1, 1, 512)    262144      add_28[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_85 (BatchNo (None, 1, 1, 512)    2048        conv2d_90[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_85 (Activation)      (None, 1, 1, 512)    0           batch_normalization_85[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_91 (Conv2D)              (None, 1, 1, 512)    2359296     activation_85[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_86 (BatchNo (None, 1, 1, 512)    2048        conv2d_91[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_86 (Activation)      (None, 1, 1, 512)    0           batch_normalization_86[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_92 (Conv2D)              (None, 1, 1, 512)    262144      activation_86[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_87 (BatchNo (None, 1, 1, 512)    2048        conv2d_92[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_87 (Activation)      (None, 1, 1, 512)    0           batch_normalization_87[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_29 (Gl (None, 512)          0           activation_87[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_29 (Reshape)            (None, 1, 1, 512)    0           global_average_pooling2d_29[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_57 (Dense)                (None, 1, 1, 32)     16384       reshape_29[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_58 (Dense)                (None, 1, 1, 512)    16384       dense_57[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_29 (Multiply)          (None, 1, 1, 512)    0           activation_87[0][0]              \n",
      "                                                                 dense_58[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_29 (Add)                    (None, 1, 1, 512)    0           multiply_29[0][0]                \n",
      "                                                                 add_28[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_93 (Conv2D)              (None, 1, 1, 512)    262144      add_29[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_88 (BatchNo (None, 1, 1, 512)    2048        conv2d_93[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_88 (Activation)      (None, 1, 1, 512)    0           batch_normalization_88[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_94 (Conv2D)              (None, 1, 1, 512)    2359296     activation_88[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_89 (BatchNo (None, 1, 1, 512)    2048        conv2d_94[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_89 (Activation)      (None, 1, 1, 512)    0           batch_normalization_89[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_95 (Conv2D)              (None, 1, 1, 512)    262144      activation_89[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_90 (BatchNo (None, 1, 1, 512)    2048        conv2d_95[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_90 (Activation)      (None, 1, 1, 512)    0           batch_normalization_90[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_30 (Gl (None, 512)          0           activation_90[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_30 (Reshape)            (None, 1, 1, 512)    0           global_average_pooling2d_30[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_59 (Dense)                (None, 1, 1, 32)     16384       reshape_30[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_60 (Dense)                (None, 1, 1, 512)    16384       dense_59[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_30 (Multiply)          (None, 1, 1, 512)    0           activation_90[0][0]              \n",
      "                                                                 dense_60[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_30 (Add)                    (None, 1, 1, 512)    0           multiply_30[0][0]                \n",
      "                                                                 add_29[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_96 (Conv2D)              (None, 1, 1, 512)    262144      add_30[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_91 (BatchNo (None, 1, 1, 512)    2048        conv2d_96[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_91 (Activation)      (None, 1, 1, 512)    0           batch_normalization_91[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_97 (Conv2D)              (None, 1, 1, 512)    2359296     activation_91[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_92 (BatchNo (None, 1, 1, 512)    2048        conv2d_97[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_92 (Activation)      (None, 1, 1, 512)    0           batch_normalization_92[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_98 (Conv2D)              (None, 1, 1, 512)    262144      activation_92[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_93 (BatchNo (None, 1, 1, 512)    2048        conv2d_98[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_93 (Activation)      (None, 1, 1, 512)    0           batch_normalization_93[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_31 (Gl (None, 512)          0           activation_93[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_31 (Reshape)            (None, 1, 1, 512)    0           global_average_pooling2d_31[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_61 (Dense)                (None, 1, 1, 32)     16384       reshape_31[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_62 (Dense)                (None, 1, 1, 512)    16384       dense_61[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_31 (Multiply)          (None, 1, 1, 512)    0           activation_93[0][0]              \n",
      "                                                                 dense_62[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_31 (Add)                    (None, 1, 1, 512)    0           multiply_31[0][0]                \n",
      "                                                                 add_30[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_99 (Conv2D)              (None, 1, 1, 512)    262144      add_31[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_94 (BatchNo (None, 1, 1, 512)    2048        conv2d_99[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_94 (Activation)      (None, 1, 1, 512)    0           batch_normalization_94[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_100 (Conv2D)             (None, 1, 1, 512)    2359296     activation_94[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_95 (BatchNo (None, 1, 1, 512)    2048        conv2d_100[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "activation_95 (Activation)      (None, 1, 1, 512)    0           batch_normalization_95[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_101 (Conv2D)             (None, 1, 1, 512)    262144      activation_95[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_96 (BatchNo (None, 1, 1, 512)    2048        conv2d_101[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "activation_96 (Activation)      (None, 1, 1, 512)    0           batch_normalization_96[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_32 (Gl (None, 512)          0           activation_96[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "reshape_32 (Reshape)            (None, 1, 1, 512)    0           global_average_pooling2d_32[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "dense_63 (Dense)                (None, 1, 1, 32)     16384       reshape_32[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "dense_64 (Dense)                (None, 1, 1, 512)    16384       dense_63[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "multiply_32 (Multiply)          (None, 1, 1, 512)    0           activation_96[0][0]              \n",
      "                                                                 dense_64[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "add_32 (Add)                    (None, 1, 1, 512)    0           multiply_32[0][0]                \n",
      "                                                                 add_31[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_33 (Gl (None, 512)          0           add_32[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "dense_65 (Dense)                (None, 10)           5120        global_average_pooling2d_33[0][0]\n",
      "==================================================================================================\n",
      "Total params: 28,253,760\n",
      "Trainable params: 28,208,448\n",
      "Non-trainable params: 45,312\n",
      "__________________________________________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = create_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Basic Training code"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model checkpoint file path: ./1-20181109-222507.hdf5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/keras/callbacks.py:999: UserWarning: `epsilon` argument is deprecated and will be removed, use `min_delta` instead.\n",
      "  warnings.warn('`epsilon` argument is deprecated and '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/100\n",
      "60000/60000 [==============================] - 199s 3ms/step - loss: 1.7666 - acc: 0.7969 - val_loss: 0.9411 - val_acc: 0.8985\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.94109, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 2/100\n",
      "60000/60000 [==============================] - 182s 3ms/step - loss: 1.2153 - acc: 0.8685 - val_loss: 1.1228 - val_acc: 0.8953\n",
      "\n",
      "Epoch 00002: val_loss did not improve from 0.94109\n",
      "Epoch 3/100\n",
      "60000/60000 [==============================] - 185s 3ms/step - loss: 0.9359 - acc: 0.9009 - val_loss: 1.7641 - val_acc: 0.7658\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 0.94109\n",
      "\n",
      "Epoch 00003: ReduceLROnPlateau reducing learning rate to 0.00020000000949949026.\n",
      "Epoch 4/100\n",
      "60000/60000 [==============================] - 189s 3ms/step - loss: 0.7529 - acc: 0.9255 - val_loss: 0.7374 - val_acc: 0.9343\n",
      "\n",
      "Epoch 00004: val_loss improved from 0.94109 to 0.73740, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 5/100\n",
      "60000/60000 [==============================] - 192s 3ms/step - loss: 0.8241 - acc: 0.9280 - val_loss: 0.7556 - val_acc: 0.9346\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.73740\n",
      "Epoch 6/100\n",
      "60000/60000 [==============================] - 193s 3ms/step - loss: 0.7146 - acc: 0.9387 - val_loss: 0.7553 - val_acc: 0.9369.71 - ETA: 4s - loss: 0. - ETA: 2s\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.73740\n",
      "\n",
      "Epoch 00006: ReduceLROnPlateau reducing learning rate to 4.0000001899898055e-05.\n",
      "Epoch 7/100\n",
      "60000/60000 [==============================] - 194s 3ms/step - loss: 0.6400 - acc: 0.9481 - val_loss: 0.6045 - val_acc: 0.9472\n",
      "\n",
      "Epoch 00007: val_loss improved from 0.73740 to 0.60451, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 8/100\n",
      "60000/60000 [==============================] - 193s 3ms/step - loss: 0.5627 - acc: 0.9543 - val_loss: 0.5852 - val_acc: 0.9491\n",
      "\n",
      "Epoch 00008: val_loss improved from 0.60451 to 0.58521, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 9/100\n",
      "60000/60000 [==============================] - 194s 3ms/step - loss: 0.5603 - acc: 0.9552 - val_loss: 0.5829 - val_acc: 0.9500\n",
      "\n",
      "Epoch 00009: val_loss improved from 0.58521 to 0.58295, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 10/100\n",
      "60000/60000 [==============================] - 195s 3ms/step - loss: 0.5677 - acc: 0.9548 - val_loss: 0.5804 - val_acc: 0.9489\n",
      "\n",
      "Epoch 00010: val_loss improved from 0.58295 to 0.58042, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 11/100\n",
      "60000/60000 [==============================] - 195s 3ms/step - loss: 0.4915 - acc: 0.9604 - val_loss: 0.5623 - val_acc: 0.9511\n",
      "\n",
      "Epoch 00011: val_loss improved from 0.58042 to 0.56226, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 12/100\n",
      "60000/60000 [==============================] - 193s 3ms/step - loss: 0.4836 - acc: 0.9614 - val_loss: 0.4900 - val_acc: 0.9570\n",
      "\n",
      "Epoch 00012: val_loss improved from 0.56226 to 0.48997, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 13/100\n",
      "60000/60000 [==============================] - 192s 3ms/step - loss: 0.4619 - acc: 0.9627 - val_loss: 0.4846 - val_acc: 0.9548\n",
      "\n",
      "Epoch 00013: val_loss improved from 0.48997 to 0.48463, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 14/100\n",
      "60000/60000 [==============================] - 186s 3ms/step - loss: 0.5313 - acc: 0.9597 - val_loss: 0.5997 - val_acc: 0.9495\n",
      "\n",
      "Epoch 00014: val_loss did not improve from 0.48463\n",
      "Epoch 15/100\n",
      "60000/60000 [==============================] - 195s 3ms/step - loss: 0.4781 - acc: 0.9629 - val_loss: 0.4800 - val_acc: 0.9572\n",
      "\n",
      "Epoch 00015: val_loss improved from 0.48463 to 0.48000, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 16/100\n",
      "60000/60000 [==============================] - 194s 3ms/step - loss: 0.4312 - acc: 0.9662 - val_loss: 0.4756 - val_acc: 0.9575\n",
      "\n",
      "Epoch 00016: val_loss improved from 0.48000 to 0.47559, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 17/100\n",
      "60000/60000 [==============================] - 193s 3ms/step - loss: 0.4270 - acc: 0.9669 - val_loss: 0.5129 - val_acc: 0.9555s: 0.4282 - acc:\n",
      "\n",
      "Epoch 00017: val_loss did not improve from 0.47559\n",
      "Epoch 18/100\n",
      "60000/60000 [==============================] - 192s 3ms/step - loss: 0.4010 - acc: 0.9687 - val_loss: 0.4485 - val_acc: 0.9593\n",
      "\n",
      "Epoch 00018: val_loss improved from 0.47559 to 0.44846, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 19/100\n",
      "60000/60000 [==============================] - 194s 3ms/step - loss: 0.3879 - acc: 0.9702 - val_loss: 0.4366 - val_acc: 0.9609\n",
      "\n",
      "Epoch 00019: val_loss improved from 0.44846 to 0.43660, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 20/100\n",
      "60000/60000 [==============================] - 187s 3ms/step - loss: 0.3922 - acc: 0.9695 - val_loss: 0.4138 - val_acc: 0.9611\n",
      "\n",
      "Epoch 00020: val_loss improved from 0.43660 to 0.41377, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 21/100\n",
      "60000/60000 [==============================] - 188s 3ms/step - loss: 0.3544 - acc: 0.9726 - val_loss: 0.3830 - val_acc: 0.9641\n",
      "\n",
      "Epoch 00021: val_loss improved from 0.41377 to 0.38304, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 22/100\n",
      "60000/60000 [==============================] - 187s 3ms/step - loss: 0.3530 - acc: 0.9729 - val_loss: 0.4459 - val_acc: 0.9599\n",
      "\n",
      "Epoch 00022: val_loss did not improve from 0.38304\n",
      "Epoch 23/100\n",
      "60000/60000 [==============================] - 187s 3ms/step - loss: 0.3339 - acc: 0.9740 - val_loss: 0.4011 - val_acc: 0.9622\n",
      "\n",
      "Epoch 00023: val_loss did not improve from 0.38304\n",
      "\n",
      "Epoch 00023: ReduceLROnPlateau reducing learning rate to 8.000000525498762e-06.\n",
      "Epoch 24/100\n",
      "60000/60000 [==============================] - 187s 3ms/step - loss: 0.2945 - acc: 0.9771 - val_loss: 0.3434 - val_acc: 0.9668\n",
      "\n",
      "Epoch 00024: val_loss improved from 0.38304 to 0.34342, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 25/100\n",
      "60000/60000 [==============================] - 196s 3ms/step - loss: 0.2701 - acc: 0.9798 - val_loss: 0.3480 - val_acc: 0.9666\n",
      "\n",
      "Epoch 00025: val_loss did not improve from 0.34342\n",
      "Epoch 26/100\n",
      "60000/60000 [==============================] - 213s 4ms/step - loss: 0.2664 - acc: 0.9802 - val_loss: 0.3253 - val_acc: 0.9683\n",
      "\n",
      "Epoch 00026: val_loss improved from 0.34342 to 0.32528, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 27/100\n",
      "60000/60000 [==============================] - 209s 3ms/step - loss: 0.2610 - acc: 0.9806 - val_loss: 0.3344 - val_acc: 0.9688\n",
      "\n",
      "Epoch 00027: val_loss did not improve from 0.32528\n",
      "Epoch 28/100\n",
      "60000/60000 [==============================] - 209s 3ms/step - loss: 0.2628 - acc: 0.9805 - val_loss: 0.3179 - val_acc: 0.9687\n",
      "\n",
      "Epoch 00028: val_loss improved from 0.32528 to 0.31794, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 29/100\n",
      "60000/60000 [==============================] - 209s 3ms/step - loss: 0.2538 - acc: 0.9809 - val_loss: 0.3109 - val_acc: 0.9689\n",
      "\n",
      "Epoch 00029: val_loss improved from 0.31794 to 0.31088, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 30/100\n",
      "60000/60000 [==============================] - 210s 3ms/step - loss: 0.2440 - acc: 0.9819 - val_loss: 0.3096 - val_acc: 0.9695\n",
      "\n",
      "Epoch 00030: val_loss improved from 0.31088 to 0.30961, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 31/100\n",
      "60000/60000 [==============================] - 209s 3ms/step - loss: 0.2395 - acc: 0.9822 - val_loss: 0.3121 - val_acc: 0.9702\n",
      "\n",
      "Epoch 00031: val_loss did not improve from 0.30961\n",
      "Epoch 32/100\n",
      "60000/60000 [==============================] - 209s 3ms/step - loss: 0.2326 - acc: 0.9823 - val_loss: 0.3019 - val_acc: 0.9700\n",
      "\n",
      "Epoch 00032: val_loss improved from 0.30961 to 0.30185, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 33/100\n",
      "60000/60000 [==============================] - 210s 3ms/step - loss: 0.2324 - acc: 0.9825 - val_loss: 0.3012 - val_acc: 0.9694\n",
      "\n",
      "Epoch 00033: val_loss improved from 0.30185 to 0.30123, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 34/100\n",
      "60000/60000 [==============================] - 209s 3ms/step - loss: 0.2318 - acc: 0.9831 - val_loss: 0.3071 - val_acc: 0.9703\n",
      "\n",
      "Epoch 00034: val_loss did not improve from 0.30123\n",
      "Epoch 35/100\n",
      "60000/60000 [==============================] - 210s 4ms/step - loss: 0.2243 - acc: 0.9833 - val_loss: 0.3014 - val_acc: 0.9706\n",
      "\n",
      "Epoch 00035: val_loss did not improve from 0.30123\n",
      "\n",
      "Epoch 00035: ReduceLROnPlateau reducing learning rate to 1.6000001778593287e-06.\n",
      "Epoch 36/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000/60000 [==============================] - 207s 3ms/step - loss: 0.2195 - acc: 0.9841 - val_loss: 0.3000 - val_acc: 0.9702\n",
      "\n",
      "Epoch 00036: val_loss improved from 0.30123 to 0.29998, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 37/100\n",
      "60000/60000 [==============================] - 207s 3ms/step - loss: 0.2145 - acc: 0.9841 - val_loss: 0.2983 - val_acc: 0.9698\n",
      "\n",
      "Epoch 00037: val_loss improved from 0.29998 to 0.29827, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 38/100\n",
      "60000/60000 [==============================] - 206s 3ms/step - loss: 0.2124 - acc: 0.9841 - val_loss: 0.2956 - val_acc: 0.9706\n",
      "\n",
      "Epoch 00038: val_loss improved from 0.29827 to 0.29561, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 39/100\n",
      "60000/60000 [==============================] - 206s 3ms/step - loss: 0.2103 - acc: 0.9847 - val_loss: 0.2947 - val_acc: 0.9709\n",
      "\n",
      "Epoch 00039: val_loss improved from 0.29561 to 0.29472, saving model to ./1-20181109-222507.hdf5\n",
      "Epoch 40/100\n",
      "60000/60000 [==============================] - 206s 3ms/step - loss: 0.2096 - acc: 0.9845 - val_loss: 0.3011 - val_acc: 0.9702\n",
      "\n",
      "Epoch 00040: val_loss did not improve from 0.29472\n",
      "Epoch 41/100\n",
      "60000/60000 [==============================] - 206s 3ms/step - loss: 0.2097 - acc: 0.9847 - val_loss: 0.3016 - val_acc: 0.9700\n",
      "\n",
      "Epoch 00041: val_loss did not improve from 0.29472\n",
      "\n",
      "Epoch 00041: ReduceLROnPlateau reducing learning rate to 3.200000264769187e-07.\n",
      "Epoch 42/100\n",
      "60000/60000 [==============================] - 206s 3ms/step - loss: 0.2043 - acc: 0.9848 - val_loss: 0.2998 - val_acc: 0.9699\n",
      "\n",
      "Epoch 00042: val_loss did not improve from 0.29472\n",
      "Epoch 00042: early stopping\n",
      "model training complete. time spent: 2:20:40.228941\n"
     ]
    }
   ],
   "source": [
    "#model_filename:where the model is checkpointed\n",
    "model_id = 1\n",
    "model_dir = './'\n",
    "timestr = time.strftime(\"%Y%m%d-%H%M%S\")\n",
    "model_filename = model_dir + '{}-{}.hdf5'.format(model_id, timestr)\n",
    "print('model checkpoint file path: {}'.format(model_filename))\n",
    "lr_reduction_factor = 0.2\n",
    "min_learning_rate = 1e-07\n",
    "#Adding early stopping,model_checkpoint,reduceLRonPlateau\n",
    "early_stop = EarlyStopping(monitor='val_loss',\n",
    "                           patience=3,\n",
    "                           min_delta=0, \n",
    "                           verbose=1,\n",
    "                           mode='auto')\n",
    "\n",
    "model_checkpoint = ModelCheckpoint(model_filename,\n",
    "                                   monitor='val_loss',\n",
    "                                   verbose=1,\n",
    "                                   save_best_only=True)\n",
    "\n",
    "reduceLR = ReduceLROnPlateau(monitor='val_loss',\n",
    "                             factor=lr_reduction_factor,\n",
    "                             patience=2,\n",
    "                             verbose=1,\n",
    "                             min_lr=min_learning_rate,\n",
    "                             epsilon=1e-4)\n",
    "training_start_time = datetime.now()\n",
    "model.compile(optimizer=Adam(0.001),loss=\"categorical_crossentropy\",metrics=[\"accuracy\"])\n",
    "history = model.fit(train_x,train_y,batch_size=32,epochs=100,verbose=1,callbacks=[model_checkpoint, early_stop, reduceLR],validation_data=(test_x, test_y))\n",
    "time_spent_trianing = datetime.now() - training_start_time\n",
    "print('model training complete. time spent: {}'.format(time_spent_trianing)) time spent: {}'.format(time_spent_trianing))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Printing the training statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'acc': [0.7968666666666666, 0.8685, 0.9009, 0.9254666666666667, 0.92795, 0.9386833333333333, 0.9480666666666666, 0.9542833333333334, 0.9552333333333334, 0.95475, 0.9604166666666667, 0.9613666666666667, 0.9627333333333333, 0.9596833333333333, 0.9629, 0.9662333333333334, 0.96695, 0.9687333333333333, 0.9702333333333333, 0.96945, 0.9725666666666667, 0.9728833333333333, 0.9740333333333333, 0.9771333333333333, 0.97975, 0.9801666666666666, 0.9806166666666667, 0.9805333333333334, 0.98095, 0.9819333333333333, 0.98225, 0.9822666666666666, 0.9825333333333334, 0.9830833333333333, 0.9833166666666666, 0.9840666666666666, 0.9841333333333333, 0.98415, 0.9846666666666667, 0.9844666666666667, 0.9847333333333333, 0.9848333333333333], 'loss': [1.766570236279567, 1.2153313704480728, 0.9359473824123542, 0.7528933323204517, 0.8241054387420416, 0.7145960925117135, 0.6399572514941295, 0.56269377814581, 0.5603003170053165, 0.5677226155792674, 0.49146803734799227, 0.4836316032325228, 0.4618825278525551, 0.5313111084590355, 0.47810490656544763, 0.4311915253142516, 0.4269553929189841, 0.40097368479917445, 0.387906656554838, 0.39215852171430987, 0.35437056815425555, 0.35297940527598065, 0.3339343698233366, 0.2944901867126425, 0.270148833497862, 0.26643246626357237, 0.2610493294859926, 0.2628087000812093, 0.2538131102204323, 0.24401601050148408, 0.2394869392901659, 0.23261970695306858, 0.23241221536695958, 0.23176814110577107, 0.22431538757284483, 0.21946482640703519, 0.2145123631219069, 0.2123597050865491, 0.21033813654432695, 0.2095932347153624, 0.20973196470886468, 0.20432266338268915], 'val_acc': [0.8985, 0.8953, 0.7658, 0.9343, 0.9346, 0.9369, 0.9472, 0.9491, 0.95, 0.9489, 0.9511, 0.957, 0.9548, 0.9495, 0.9572, 0.9575, 0.9555, 0.9593, 0.9609, 0.9611, 0.9641, 0.9599, 0.9622, 0.9668, 0.9666, 0.9683, 0.9688, 0.9687, 0.9689, 0.9695, 0.9702, 0.97, 0.9694, 0.9703, 0.9706, 0.9702, 0.9698, 0.9706, 0.9709, 0.9702, 0.97, 0.9699], 'val_loss': [0.9410888967633247, 1.1228376137286424, 1.7641196365237235, 0.7374023982107639, 0.7556161650806665, 0.7553232158184051, 0.604510620841384, 0.5852146054938435, 0.5829487362638116, 0.5804187469005585, 0.5622595959484578, 0.4899738857835531, 0.48462682275772095, 0.599685810445249, 0.47999777824729684, 0.47558509978204966, 0.5129130979776383, 0.4484583640620112, 0.43660329358577726, 0.41377155751287936, 0.38304409222900865, 0.44586758334487675, 0.40109450864046814, 0.34341568670719863, 0.34802285021543505, 0.32528242510557176, 0.3343873058080673, 0.3179416278496385, 0.3108842324092984, 0.30960643691718576, 0.31212934133708475, 0.30185063314288857, 0.30122580584734676, 0.3071012426301837, 0.3014180638372898, 0.2999803487062454, 0.29826539500802757, 0.2956123667880893, 0.29472493720948695, 0.3011151848882437, 0.3015569223999977, 0.29975422763973475], 'lr': [0.001, 0.001, 0.001, 0.00020000001, 0.00020000001, 0.00020000001, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 4.0000003e-05, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 8.000001e-06, 1.6000001e-06, 1.6000001e-06, 1.6000001e-06, 1.6000001e-06, 1.6000001e-06, 1.6000001e-06, 3.2000003e-07]}\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x720 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pickle\n",
    "print(history.history)\n",
    "\n",
    "historyFilePath = model_dir + '{}-{}-train-history.png'.format(model_id, timestr)\n",
    "trainingHistoryPlot(str(model_id) + str(timestr), historyFilePath, history.history)\n",
    "\n",
    "pickleFilePath = model_dir + '{}-{}-history-dict.pickle'.format(model_id, timestr)\n",
    "with open(pickleFilePath, 'wb') as handle:\n",
    "    pickle.dump(history.history, handle, protocol=pickle.HIGHEST_PROTOCOL)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Loading the model and evaluating Test set accuracy again"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy:  0.9709\n"
     ]
    }
   ],
   "source": [
    "model_filename = './1-20181109-222507.hdf5'\n",
    "model_id = 1\n",
    "model_dir = './'\n",
    "model.load_weights(model_filename)\n",
    "model.compile(optimizer=Adam(0.001),loss=\"categorical_crossentropy\",metrics=[\"accuracy\"])\n",
    "score = model.evaluate(test_x, test_y, verbose=0)\n",
    "print('Accuracy: ',score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Got a test accuracy of around 97%. Accuracy can be improved by adding methods such as cutout, data augmentation and distortion."
   ]
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}