train_model_from_MobileNet.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import keras\n",
    "from keras import backend as K\n",
    "from keras.models import Sequential,load_model,Model\n",
    "from keras.layers.core import Dense, Flatten\n",
    "from keras.optimizers import Adam\n",
    "from keras.metrics import categorical_crossentropy\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.convolutional import *\n",
    "from matplotlib import pyplot as plt\n",
    "import itertools\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_path = 'data/train'\n",
    "test_path = 'data/test'\n",
    "valid_path = 'data/valid'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 1200 images belonging to 6 classes.\n",
      "Found 60 images belonging to 6 classes.\n",
      "Found 251 images belonging to 6 classes.\n"
     ]
    }
   ],
   "source": [
    "train_batches = ImageDataGenerator(rescale=1./255,zoom_range=0.2).flow_from_directory(train_path,target_size=(224,224),classes=['0','1','2','3','4','5'],batch_size=10) \n",
    "test_batches = ImageDataGenerator(rescale=1./255,zoom_range=0.2).flow_from_directory(test_path,target_size= (224,224),classes=['0','1','2','3','4','5'],batch_size=5)\n",
    "valid_batches = ImageDataGenerator(rescale=1./255,zoom_range=0.2).flow_from_directory(valid_path,target_size=(224,224),classes=['0','1','2','3','4','5'],batch_size=10)\n",
    "# train_batches = ImageDataGenerator().flow_from_directory(train_path,target_size=(224,224),classes=['-1','0','1','2','3','4'],batch_size=10) \n",
    "# test_batches = ImageDataGenerator().flow_from_directory(test_path,target_size= (224,224),classes=['-1','0','1','2','3','4'],batch_size=5)\n",
    "# valid_batches = ImageDataGenerator().flow_from_directory(valid_path,target_size=(224,224),classes=['-1','0','1','2','3','4'],batch_size=10)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plots(ims,figsize=(12,6),rows=1,interp=False,titles=None):\n",
    "    if type(ims[0]) is np.ndarray:\n",
    "        ims = np.array(ims).astype(np.uint8)\n",
    "        if (ims.shape[-1]!=3):\n",
    "            ims = ims.transpose((0,2,3,1))\n",
    "    f = plt.figure(figsize=figsize)\n",
    "    cols = len(ims)//rows if len(ims) % 2 ==0 else len(ims)//rows + 1\n",
    "    for i in range(len(ims)):\n",
    "        sp = f.add_subplot(rows,cols,i+1)\n",
    "        sp.axis('Off')\n",
    "        if titles is not None:\n",
    "            sp.set_title(titles[i],fontsize=16)\n",
    "        plt.imshow(ims[i],interpolation=None if interp else 'none')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "imgs, labels =next(train_batches)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Anaconda3\\lib\\site-packages\\matplotlib\\text.py:1190: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison\n",
      "  if s != self._text:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 10 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plots(imgs,titles=labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "mobile = keras.applications.mobilenet.MobileNet()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 224, 224, 3)       0         \n",
      "_________________________________________________________________\n",
      "conv1_pad (ZeroPadding2D)    (None, 226, 226, 3)       0         \n",
      "_________________________________________________________________\n",
      "conv1 (Conv2D)               (None, 112, 112, 32)      864       \n",
      "_________________________________________________________________\n",
      "conv1_bn (BatchNormalization (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv1_relu (Activation)      (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_pad_1 (ZeroPadding2D)   (None, 114, 114, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_1 (DepthwiseConv2D)  (None, 112, 112, 32)      288       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_bn (BatchNormaliza (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_relu (Activation)  (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_pw_1 (Conv2D)           (None, 112, 112, 64)      2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_1_bn (BatchNormaliza (None, 112, 112, 64)      256       \n",
      "_________________________________________________________________\n",
      "conv_pw_1_relu (Activation)  (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_pad_2 (ZeroPadding2D)   (None, 114, 114, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_2 (DepthwiseConv2D)  (None, 56, 56, 64)        576       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_bn (BatchNormaliza (None, 56, 56, 64)        256       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_relu (Activation)  (None, 56, 56, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_2 (Conv2D)           (None, 56, 56, 128)       8192      \n",
      "_________________________________________________________________\n",
      "conv_pw_2_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_2_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_3 (ZeroPadding2D)   (None, 58, 58, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_3 (DepthwiseConv2D)  (None, 56, 56, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_3_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_3 (Conv2D)           (None, 56, 56, 128)       16384     \n",
      "_________________________________________________________________\n",
      "conv_pw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_3_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_4 (ZeroPadding2D)   (None, 58, 58, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_4 (DepthwiseConv2D)  (None, 28, 28, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_4_bn (BatchNormaliza (None, 28, 28, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_4_relu (Activation)  (None, 28, 28, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_4 (Conv2D)           (None, 28, 28, 256)       32768     \n",
      "_________________________________________________________________\n",
      "conv_pw_4_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_4_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_5 (ZeroPadding2D)   (None, 30, 30, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_5 (DepthwiseConv2D)  (None, 28, 28, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_5 (Conv2D)           (None, 28, 28, 256)       65536     \n",
      "_________________________________________________________________\n",
      "conv_pw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_5_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_6 (ZeroPadding2D)   (None, 30, 30, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_6 (DepthwiseConv2D)  (None, 14, 14, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_bn (BatchNormaliza (None, 14, 14, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_relu (Activation)  (None, 14, 14, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_6 (Conv2D)           (None, 14, 14, 512)       131072    \n",
      "_________________________________________________________________\n",
      "conv_pw_6_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_6_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_7 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_7 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_7 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_7_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_8 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_8 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_8 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_8_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_9 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_9 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_9 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_9_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_10 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_10 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_10 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_10_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_11 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_11 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_11 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_11_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_12 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_12 (DepthwiseConv2D) (None, 7, 7, 512)         4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_bn (BatchNormaliz (None, 7, 7, 512)         2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_relu (Activation) (None, 7, 7, 512)         0         \n",
      "_________________________________________________________________\n",
      "conv_pw_12 (Conv2D)          (None, 7, 7, 1024)        524288    \n",
      "_________________________________________________________________\n",
      "conv_pw_12_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_12_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_pad_13 (ZeroPadding2D)  (None, 9, 9, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_dw_13 (DepthwiseConv2D) (None, 7, 7, 1024)        9216      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_13 (Conv2D)          (None, 7, 7, 1024)        1048576   \n",
      "_________________________________________________________________\n",
      "conv_pw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_13_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_2 ( (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "reshape_1 (Reshape)          (None, 1, 1, 1024)        0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 1, 1, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_preds (Conv2D)          (None, 1, 1, 1000)        1025000   \n",
      "_________________________________________________________________\n",
      "act_softmax (Activation)     (None, 1, 1, 1000)        0         \n",
      "_________________________________________________________________\n",
      "reshape_2 (Reshape)          (None, 1000)              0         \n",
      "=================================================================\n",
      "Total params: 4,253,864\n",
      "Trainable params: 4,231,976\n",
      "Non-trainable params: 21,888\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "mobile.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "x= mobile.layers[-6].output\n",
    "predictions = Dense(6,activation='softmax')(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Model(inputs=mobile.input,outputs=predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 224, 224, 3)       0         \n",
      "_________________________________________________________________\n",
      "conv1_pad (ZeroPadding2D)    (None, 226, 226, 3)       0         \n",
      "_________________________________________________________________\n",
      "conv1 (Conv2D)               (None, 112, 112, 32)      864       \n",
      "_________________________________________________________________\n",
      "conv1_bn (BatchNormalization (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv1_relu (Activation)      (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_pad_1 (ZeroPadding2D)   (None, 114, 114, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_1 (DepthwiseConv2D)  (None, 112, 112, 32)      288       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_bn (BatchNormaliza (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_relu (Activation)  (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_pw_1 (Conv2D)           (None, 112, 112, 64)      2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_1_bn (BatchNormaliza (None, 112, 112, 64)      256       \n",
      "_________________________________________________________________\n",
      "conv_pw_1_relu (Activation)  (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_pad_2 (ZeroPadding2D)   (None, 114, 114, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_2 (DepthwiseConv2D)  (None, 56, 56, 64)        576       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_bn (BatchNormaliza (None, 56, 56, 64)        256       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_relu (Activation)  (None, 56, 56, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_2 (Conv2D)           (None, 56, 56, 128)       8192      \n",
      "_________________________________________________________________\n",
      "conv_pw_2_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_2_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_3 (ZeroPadding2D)   (None, 58, 58, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_3 (DepthwiseConv2D)  (None, 56, 56, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_3_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_3 (Conv2D)           (None, 56, 56, 128)       16384     \n",
      "_________________________________________________________________\n",
      "conv_pw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_3_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_4 (ZeroPadding2D)   (None, 58, 58, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_4 (DepthwiseConv2D)  (None, 28, 28, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_4_bn (BatchNormaliza (None, 28, 28, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_4_relu (Activation)  (None, 28, 28, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_4 (Conv2D)           (None, 28, 28, 256)       32768     \n",
      "_________________________________________________________________\n",
      "conv_pw_4_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_4_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_5 (ZeroPadding2D)   (None, 30, 30, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_5 (DepthwiseConv2D)  (None, 28, 28, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_5 (Conv2D)           (None, 28, 28, 256)       65536     \n",
      "_________________________________________________________________\n",
      "conv_pw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_5_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_6 (ZeroPadding2D)   (None, 30, 30, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_6 (DepthwiseConv2D)  (None, 14, 14, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_bn (BatchNormaliza (None, 14, 14, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_relu (Activation)  (None, 14, 14, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_6 (Conv2D)           (None, 14, 14, 512)       131072    \n",
      "_________________________________________________________________\n",
      "conv_pw_6_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_6_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_7 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_7 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_7 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_7_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_8 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_8 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_8 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_8_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_9 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_9 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_9 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_9_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_10 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_10 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_10 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_10_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_11 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_11 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_11 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_11_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_12 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_12 (DepthwiseConv2D) (None, 7, 7, 512)         4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_bn (BatchNormaliz (None, 7, 7, 512)         2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_relu (Activation) (None, 7, 7, 512)         0         \n",
      "_________________________________________________________________\n",
      "conv_pw_12 (Conv2D)          (None, 7, 7, 1024)        524288    \n",
      "_________________________________________________________________\n",
      "conv_pw_12_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_12_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_pad_13 (ZeroPadding2D)  (None, 9, 9, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_dw_13 (DepthwiseConv2D) (None, 7, 7, 1024)        9216      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_13 (Conv2D)          (None, 7, 7, 1024)        1048576   \n",
      "_________________________________________________________________\n",
      "conv_pw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_13_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_2 ( (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 6)                 6150      \n",
      "=================================================================\n",
      "Total params: 3,235,014\n",
      "Trainable params: 3,213,126\n",
      "Non-trainable params: 21,888\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "for layer in model.layers[:-23]:\n",
    "    layer.trainable = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 224, 224, 3)       0         \n",
      "_________________________________________________________________\n",
      "conv1_pad (ZeroPadding2D)    (None, 226, 226, 3)       0         \n",
      "_________________________________________________________________\n",
      "conv1 (Conv2D)               (None, 112, 112, 32)      864       \n",
      "_________________________________________________________________\n",
      "conv1_bn (BatchNormalization (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv1_relu (Activation)      (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_pad_1 (ZeroPadding2D)   (None, 114, 114, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_1 (DepthwiseConv2D)  (None, 112, 112, 32)      288       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_bn (BatchNormaliza (None, 112, 112, 32)      128       \n",
      "_________________________________________________________________\n",
      "conv_dw_1_relu (Activation)  (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv_pw_1 (Conv2D)           (None, 112, 112, 64)      2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_1_bn (BatchNormaliza (None, 112, 112, 64)      256       \n",
      "_________________________________________________________________\n",
      "conv_pw_1_relu (Activation)  (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_pad_2 (ZeroPadding2D)   (None, 114, 114, 64)      0         \n",
      "_________________________________________________________________\n",
      "conv_dw_2 (DepthwiseConv2D)  (None, 56, 56, 64)        576       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_bn (BatchNormaliza (None, 56, 56, 64)        256       \n",
      "_________________________________________________________________\n",
      "conv_dw_2_relu (Activation)  (None, 56, 56, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_2 (Conv2D)           (None, 56, 56, 128)       8192      \n",
      "_________________________________________________________________\n",
      "conv_pw_2_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_2_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_3 (ZeroPadding2D)   (None, 58, 58, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_3 (DepthwiseConv2D)  (None, 56, 56, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_3_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_3 (Conv2D)           (None, 56, 56, 128)       16384     \n",
      "_________________________________________________________________\n",
      "conv_pw_3_bn (BatchNormaliza (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_pw_3_relu (Activation)  (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_4 (ZeroPadding2D)   (None, 58, 58, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_4 (DepthwiseConv2D)  (None, 28, 28, 128)       1152      \n",
      "_________________________________________________________________\n",
      "conv_dw_4_bn (BatchNormaliza (None, 28, 28, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv_dw_4_relu (Activation)  (None, 28, 28, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_4 (Conv2D)           (None, 28, 28, 256)       32768     \n",
      "_________________________________________________________________\n",
      "conv_pw_4_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_4_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_5 (ZeroPadding2D)   (None, 30, 30, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_5 (DepthwiseConv2D)  (None, 28, 28, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_5_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_5 (Conv2D)           (None, 28, 28, 256)       65536     \n",
      "_________________________________________________________________\n",
      "conv_pw_5_bn (BatchNormaliza (None, 28, 28, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_pw_5_relu (Activation)  (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_6 (ZeroPadding2D)   (None, 30, 30, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_6 (DepthwiseConv2D)  (None, 14, 14, 256)       2304      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_bn (BatchNormaliza (None, 14, 14, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv_dw_6_relu (Activation)  (None, 14, 14, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_6 (Conv2D)           (None, 14, 14, 512)       131072    \n",
      "_________________________________________________________________\n",
      "conv_pw_6_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_6_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_7 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_7 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_7_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_7 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_7_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_7_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_8 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_8 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_8_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_8 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_8_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_8_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_9 (ZeroPadding2D)   (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_9 (DepthwiseConv2D)  (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_9_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_9 (Conv2D)           (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_9_bn (BatchNormaliza (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_9_relu (Activation)  (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_10 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_10 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_10_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_10 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_10_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_10_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_11 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_11 (DepthwiseConv2D) (None, 14, 14, 512)       4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_11_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pw_11 (Conv2D)          (None, 14, 14, 512)       262144    \n",
      "_________________________________________________________________\n",
      "conv_pw_11_bn (BatchNormaliz (None, 14, 14, 512)       2048      \n",
      "_________________________________________________________________\n",
      "conv_pw_11_relu (Activation) (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_pad_12 (ZeroPadding2D)  (None, 16, 16, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv_dw_12 (DepthwiseConv2D) (None, 7, 7, 512)         4608      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_bn (BatchNormaliz (None, 7, 7, 512)         2048      \n",
      "_________________________________________________________________\n",
      "conv_dw_12_relu (Activation) (None, 7, 7, 512)         0         \n",
      "_________________________________________________________________\n",
      "conv_pw_12 (Conv2D)          (None, 7, 7, 1024)        524288    \n",
      "_________________________________________________________________\n",
      "conv_pw_12_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_12_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_pad_13 (ZeroPadding2D)  (None, 9, 9, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_dw_13 (DepthwiseConv2D) (None, 7, 7, 1024)        9216      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_dw_13_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "conv_pw_13 (Conv2D)          (None, 7, 7, 1024)        1048576   \n",
      "_________________________________________________________________\n",
      "conv_pw_13_bn (BatchNormaliz (None, 7, 7, 1024)        4096      \n",
      "_________________________________________________________________\n",
      "conv_pw_13_relu (Activation) (None, 7, 7, 1024)        0         \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_2 ( (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 6)                 6150      \n",
      "=================================================================\n",
      "Total params: 3,235,014\n",
      "Trainable params: 1,868,806\n",
      "Non-trainable params: 1,366,208\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(Adam(lr=0.0001), loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/4\n",
      "22/22 [==============================] - 107s 5s/step - loss: 0.7502 - acc: 0.7773 - val_loss: 1.6687 - val_acc: 0.4750\n",
      "Epoch 2/4\n",
      "22/22 [==============================] - 101s 5s/step - loss: 0.0995 - acc: 0.9773 - val_loss: 1.0283 - val_acc: 0.6250\n",
      "Epoch 3/4\n",
      "22/22 [==============================] - 102s 5s/step - loss: 0.0367 - acc: 1.0000 - val_loss: 0.7931 - val_acc: 0.6750\n",
      "Epoch 4/4\n",
      "22/22 [==============================] - 101s 5s/step - loss: 0.0315 - acc: 0.9955 - val_loss: 0.5541 - val_acc: 0.8500\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x13ecfc7bf60>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit_generator(train_batches, steps_per_epoch=22,validation_data=valid_batches,validation_steps=4,epochs=4,verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('mobile_net_model_v2_0-5.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0., 0., 1., 0., 0., 0.],\n",
       "       [0., 1., 0., 0., 0., 0.],\n",
       "       [0., 0., 0., 0., 1., 0.],\n",
       "       [0., 0., 0., 1., 0., 0.],\n",
       "       [1., 0., 0., 0., 0., 0.]], dtype=float32)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_imgs,test_labels = next(test_batches)\n",
    "test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1.7291944e-03, 8.6463895e-03, 7.5096816e-01, 1.8653613e-01,\n",
       "        6.2815845e-03, 4.5838587e-02],\n",
       "       [3.7500719e-03, 4.4634381e-01, 1.4015207e-02, 2.7709427e-01,\n",
       "        1.0169696e-01, 1.5709957e-01],\n",
       "       [2.4357361e-04, 7.1739737e-04, 2.5514979e-03, 3.7331486e-03,\n",
       "        9.8814791e-01, 4.6064528e-03],\n",
       "       [1.2260924e-04, 6.4574979e-04, 1.9310594e-02, 9.7216189e-01,\n",
       "        9.3921152e-04, 6.8199639e-03],\n",
       "       [4.8696420e-01, 1.8654093e-02, 1.2064379e-01, 1.5871014e-01,\n",
       "        2.3399660e-02, 1.9162811e-01]], dtype=float32)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions = model.predict(test_imgs)\n",
    "predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Try load the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.utils.generic_utils import CustomObjectScope\n",
    "with CustomObjectScope({'relu6': keras.applications.mobilenet.relu6,'DepthwiseConv2D': keras.applications.mobilenet.DepthwiseConv2D}):\n",
    "    Model = load_model('mobile_net_model.h5')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}