4-1. InceptionV3 Naive KNN Model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import warnings\n",
    "warnings.simplefilter('ignore')\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import sys, os, time, gc\n",
    "import requests, shutil\n",
    "from sklearn.neighbors import NearestNeighbors\n",
    "import keras\n",
    "from keras.preprocessing.image import load_img, img_to_array\n",
    "from keras.applications.inception_v3 import InceptionV3\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Data Information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train:\t\t (113783, 4)\n",
      "Validation:\t (22255, 4)\n",
      "Test:\t\t (22391, 4)\n",
      "\n",
      "Train Landmarks:\t 14943\n",
      "Validation Landmarks:\t 7674\n",
      "Test Landmarks:\t\t 14436\n"
     ]
    }
   ],
   "source": [
    "train_df = pd.read_csv('./data/triplet/train.csv')\n",
    "val_df = pd.read_csv('./data/triplet/validation.csv')\n",
    "test_df = pd.read_csv('./data/triplet/test.csv')\n",
    "\n",
    "print('Train:\\t\\t', train_df.shape)\n",
    "print('Validation:\\t', val_df.shape)\n",
    "print('Test:\\t\\t', test_df.shape)\n",
    "\n",
    "print('\\nTrain Landmarks:\\t', len(train_df['landmark_id'].unique()))\n",
    "print('Validation Landmarks:\\t', len(val_df['landmark_id'].unique()))\n",
    "print('Test Landmarks:\\t\\t', len(test_df['landmark_id'].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image_id</th>\n",
       "      <th>id</th>\n",
       "      <th>url</th>\n",
       "      <th>landmark_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>465272</td>\n",
       "      <td>a2ccf8ed2e969f6a</td>\n",
       "      <td>https://lh4.googleusercontent.com/-TPHkS5gzvm4...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>64516</td>\n",
       "      <td>e205ca7c8dd7c027</td>\n",
       "      <td>https://lh3.googleusercontent.com/-V3RjsZtGpxE...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>928409</td>\n",
       "      <td>4e8ab93c1620e8a3</td>\n",
       "      <td>http://mw2.google.com/mw-panoramio/photos/medi...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>88809</td>\n",
       "      <td>896bf928214d1ca4</td>\n",
       "      <td>http://lh5.ggpht.com/-Cy0l41uUaGA/R--yB8vy41I/...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1001133</td>\n",
       "      <td>375d2a153bdca926</td>\n",
       "      <td>http://lh6.ggpht.com/-UqzFpnqE9bU/S_0u1RovfdI/...</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   image_id                id  \\\n",
       "0    465272  a2ccf8ed2e969f6a   \n",
       "1     64516  e205ca7c8dd7c027   \n",
       "2    928409  4e8ab93c1620e8a3   \n",
       "3     88809  896bf928214d1ca4   \n",
       "4   1001133  375d2a153bdca926   \n",
       "\n",
       "                                                 url  landmark_id  \n",
       "0  https://lh4.googleusercontent.com/-TPHkS5gzvm4...            0  \n",
       "1  https://lh3.googleusercontent.com/-V3RjsZtGpxE...            0  \n",
       "2  http://mw2.google.com/mw-panoramio/photos/medi...            0  \n",
       "3  http://lh5.ggpht.com/-Cy0l41uUaGA/R--yB8vy41I/...            0  \n",
       "4  http://lh6.ggpht.com/-UqzFpnqE9bU/S_0u1RovfdI/...            0  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Extract Features using InceptionV3 (No need to run)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 224, 224, 3)  0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 111, 111, 32) 864         input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_1 (BatchNor (None, 111, 111, 32) 96          conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_1 (Activation)       (None, 111, 111, 32) 0           batch_normalization_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 109, 109, 32) 9216        activation_1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_2 (BatchNor (None, 109, 109, 32) 96          conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_2 (Activation)       (None, 109, 109, 32) 0           batch_normalization_2[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_3 (Conv2D)               (None, 109, 109, 64) 18432       activation_2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_3 (BatchNor (None, 109, 109, 64) 192         conv2d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_3 (Activation)       (None, 109, 109, 64) 0           batch_normalization_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)  (None, 54, 54, 64)   0           activation_3[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_4 (Conv2D)               (None, 54, 54, 80)   5120        max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_4 (BatchNor (None, 54, 54, 80)   240         conv2d_4[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_4 (Activation)       (None, 54, 54, 80)   0           batch_normalization_4[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_5 (Conv2D)               (None, 52, 52, 192)  138240      activation_4[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_5 (BatchNor (None, 52, 52, 192)  576         conv2d_5[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_5 (Activation)       (None, 52, 52, 192)  0           batch_normalization_5[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2D)  (None, 25, 25, 192)  0           activation_5[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_9 (Conv2D)               (None, 25, 25, 64)   12288       max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_9 (BatchNor (None, 25, 25, 64)   192         conv2d_9[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "activation_9 (Activation)       (None, 25, 25, 64)   0           batch_normalization_9[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_7 (Conv2D)               (None, 25, 25, 48)   9216        max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_10 (Conv2D)              (None, 25, 25, 96)   55296       activation_9[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_7 (BatchNor (None, 25, 25, 48)   144         conv2d_7[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_10 (BatchNo (None, 25, 25, 96)   288         conv2d_10[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_7 (Activation)       (None, 25, 25, 48)   0           batch_normalization_7[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "activation_10 (Activation)      (None, 25, 25, 96)   0           batch_normalization_10[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_1 (AveragePoo (None, 25, 25, 192)  0           max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_6 (Conv2D)               (None, 25, 25, 64)   12288       max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_8 (Conv2D)               (None, 25, 25, 64)   76800       activation_7[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_11 (Conv2D)              (None, 25, 25, 96)   82944       activation_10[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_12 (Conv2D)              (None, 25, 25, 32)   6144        average_pooling2d_1[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_6 (BatchNor (None, 25, 25, 64)   192         conv2d_6[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_8 (BatchNor (None, 25, 25, 64)   192         conv2d_8[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_11 (BatchNo (None, 25, 25, 96)   288         conv2d_11[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_12 (BatchNo (None, 25, 25, 32)   96          conv2d_12[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_6 (Activation)       (None, 25, 25, 64)   0           batch_normalization_6[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "activation_8 (Activation)       (None, 25, 25, 64)   0           batch_normalization_8[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "activation_11 (Activation)      (None, 25, 25, 96)   0           batch_normalization_11[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_12 (Activation)      (None, 25, 25, 32)   0           batch_normalization_12[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed0 (Concatenate)            (None, 25, 25, 256)  0           activation_6[0][0]               \n",
      "                                                                 activation_8[0][0]               \n",
      "                                                                 activation_11[0][0]              \n",
      "                                                                 activation_12[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_16 (Conv2D)              (None, 25, 25, 64)   16384       mixed0[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_16 (BatchNo (None, 25, 25, 64)   192         conv2d_16[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_16 (Activation)      (None, 25, 25, 64)   0           batch_normalization_16[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_14 (Conv2D)              (None, 25, 25, 48)   12288       mixed0[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_17 (Conv2D)              (None, 25, 25, 96)   55296       activation_16[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_14 (BatchNo (None, 25, 25, 48)   144         conv2d_14[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_17 (BatchNo (None, 25, 25, 96)   288         conv2d_17[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_14 (Activation)      (None, 25, 25, 48)   0           batch_normalization_14[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_17 (Activation)      (None, 25, 25, 96)   0           batch_normalization_17[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_2 (AveragePoo (None, 25, 25, 256)  0           mixed0[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_13 (Conv2D)              (None, 25, 25, 64)   16384       mixed0[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_15 (Conv2D)              (None, 25, 25, 64)   76800       activation_14[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_18 (Conv2D)              (None, 25, 25, 96)   82944       activation_17[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_19 (Conv2D)              (None, 25, 25, 64)   16384       average_pooling2d_2[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_13 (BatchNo (None, 25, 25, 64)   192         conv2d_13[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_15 (BatchNo (None, 25, 25, 64)   192         conv2d_15[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_18 (BatchNo (None, 25, 25, 96)   288         conv2d_18[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_19 (BatchNo (None, 25, 25, 64)   192         conv2d_19[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_13 (Activation)      (None, 25, 25, 64)   0           batch_normalization_13[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_15 (Activation)      (None, 25, 25, 64)   0           batch_normalization_15[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_18 (Activation)      (None, 25, 25, 96)   0           batch_normalization_18[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_19 (Activation)      (None, 25, 25, 64)   0           batch_normalization_19[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed1 (Concatenate)            (None, 25, 25, 288)  0           activation_13[0][0]              \n",
      "                                                                 activation_15[0][0]              \n",
      "                                                                 activation_18[0][0]              \n",
      "                                                                 activation_19[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_23 (Conv2D)              (None, 25, 25, 64)   18432       mixed1[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_23 (BatchNo (None, 25, 25, 64)   192         conv2d_23[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_23 (Activation)      (None, 25, 25, 64)   0           batch_normalization_23[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_21 (Conv2D)              (None, 25, 25, 48)   13824       mixed1[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_24 (Conv2D)              (None, 25, 25, 96)   55296       activation_23[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_21 (BatchNo (None, 25, 25, 48)   144         conv2d_21[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_24 (BatchNo (None, 25, 25, 96)   288         conv2d_24[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_21 (Activation)      (None, 25, 25, 48)   0           batch_normalization_21[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_24 (Activation)      (None, 25, 25, 96)   0           batch_normalization_24[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_3 (AveragePoo (None, 25, 25, 288)  0           mixed1[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_20 (Conv2D)              (None, 25, 25, 64)   18432       mixed1[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_22 (Conv2D)              (None, 25, 25, 64)   76800       activation_21[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_25 (Conv2D)              (None, 25, 25, 96)   82944       activation_24[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_26 (Conv2D)              (None, 25, 25, 64)   18432       average_pooling2d_3[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_20 (BatchNo (None, 25, 25, 64)   192         conv2d_20[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_22 (BatchNo (None, 25, 25, 64)   192         conv2d_22[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_25 (BatchNo (None, 25, 25, 96)   288         conv2d_25[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_26 (BatchNo (None, 25, 25, 64)   192         conv2d_26[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_20 (Activation)      (None, 25, 25, 64)   0           batch_normalization_20[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_22 (Activation)      (None, 25, 25, 64)   0           batch_normalization_22[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_25 (Activation)      (None, 25, 25, 96)   0           batch_normalization_25[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_26 (Activation)      (None, 25, 25, 64)   0           batch_normalization_26[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed2 (Concatenate)            (None, 25, 25, 288)  0           activation_20[0][0]              \n",
      "                                                                 activation_22[0][0]              \n",
      "                                                                 activation_25[0][0]              \n",
      "                                                                 activation_26[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_28 (Conv2D)              (None, 25, 25, 64)   18432       mixed2[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_28 (BatchNo (None, 25, 25, 64)   192         conv2d_28[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_28 (Activation)      (None, 25, 25, 64)   0           batch_normalization_28[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_29 (Conv2D)              (None, 25, 25, 96)   55296       activation_28[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_29 (BatchNo (None, 25, 25, 96)   288         conv2d_29[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_29 (Activation)      (None, 25, 25, 96)   0           batch_normalization_29[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_27 (Conv2D)              (None, 12, 12, 384)  995328      mixed2[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_30 (Conv2D)              (None, 12, 12, 96)   82944       activation_29[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_27 (BatchNo (None, 12, 12, 384)  1152        conv2d_27[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_30 (BatchNo (None, 12, 12, 96)   288         conv2d_30[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_27 (Activation)      (None, 12, 12, 384)  0           batch_normalization_27[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_30 (Activation)      (None, 12, 12, 96)   0           batch_normalization_30[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2D)  (None, 12, 12, 288)  0           mixed2[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "mixed3 (Concatenate)            (None, 12, 12, 768)  0           activation_27[0][0]              \n",
      "                                                                 activation_30[0][0]              \n",
      "                                                                 max_pooling2d_3[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_35 (Conv2D)              (None, 12, 12, 128)  98304       mixed3[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_35 (BatchNo (None, 12, 12, 128)  384         conv2d_35[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_35 (Activation)      (None, 12, 12, 128)  0           batch_normalization_35[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_36 (Conv2D)              (None, 12, 12, 128)  114688      activation_35[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_36 (BatchNo (None, 12, 12, 128)  384         conv2d_36[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_36 (Activation)      (None, 12, 12, 128)  0           batch_normalization_36[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_32 (Conv2D)              (None, 12, 12, 128)  98304       mixed3[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_37 (Conv2D)              (None, 12, 12, 128)  114688      activation_36[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_32 (BatchNo (None, 12, 12, 128)  384         conv2d_32[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_37 (BatchNo (None, 12, 12, 128)  384         conv2d_37[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_32 (Activation)      (None, 12, 12, 128)  0           batch_normalization_32[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_37 (Activation)      (None, 12, 12, 128)  0           batch_normalization_37[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_33 (Conv2D)              (None, 12, 12, 128)  114688      activation_32[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_38 (Conv2D)              (None, 12, 12, 128)  114688      activation_37[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_33 (BatchNo (None, 12, 12, 128)  384         conv2d_33[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_38 (BatchNo (None, 12, 12, 128)  384         conv2d_38[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_33 (Activation)      (None, 12, 12, 128)  0           batch_normalization_33[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_38 (Activation)      (None, 12, 12, 128)  0           batch_normalization_38[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_4 (AveragePoo (None, 12, 12, 768)  0           mixed3[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_31 (Conv2D)              (None, 12, 12, 192)  147456      mixed3[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_34 (Conv2D)              (None, 12, 12, 192)  172032      activation_33[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_39 (Conv2D)              (None, 12, 12, 192)  172032      activation_38[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_40 (Conv2D)              (None, 12, 12, 192)  147456      average_pooling2d_4[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_31 (BatchNo (None, 12, 12, 192)  576         conv2d_31[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_34 (BatchNo (None, 12, 12, 192)  576         conv2d_34[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_39 (BatchNo (None, 12, 12, 192)  576         conv2d_39[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_40 (BatchNo (None, 12, 12, 192)  576         conv2d_40[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_31 (Activation)      (None, 12, 12, 192)  0           batch_normalization_31[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_34 (Activation)      (None, 12, 12, 192)  0           batch_normalization_34[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_39 (Activation)      (None, 12, 12, 192)  0           batch_normalization_39[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_40 (Activation)      (None, 12, 12, 192)  0           batch_normalization_40[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed4 (Concatenate)            (None, 12, 12, 768)  0           activation_31[0][0]              \n",
      "                                                                 activation_34[0][0]              \n",
      "                                                                 activation_39[0][0]              \n",
      "                                                                 activation_40[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_45 (Conv2D)              (None, 12, 12, 160)  122880      mixed4[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_45 (BatchNo (None, 12, 12, 160)  480         conv2d_45[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_45 (Activation)      (None, 12, 12, 160)  0           batch_normalization_45[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_46 (Conv2D)              (None, 12, 12, 160)  179200      activation_45[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_46 (BatchNo (None, 12, 12, 160)  480         conv2d_46[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_46 (Activation)      (None, 12, 12, 160)  0           batch_normalization_46[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_42 (Conv2D)              (None, 12, 12, 160)  122880      mixed4[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_47 (Conv2D)              (None, 12, 12, 160)  179200      activation_46[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_42 (BatchNo (None, 12, 12, 160)  480         conv2d_42[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_47 (BatchNo (None, 12, 12, 160)  480         conv2d_47[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_42 (Activation)      (None, 12, 12, 160)  0           batch_normalization_42[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_47 (Activation)      (None, 12, 12, 160)  0           batch_normalization_47[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_43 (Conv2D)              (None, 12, 12, 160)  179200      activation_42[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_48 (Conv2D)              (None, 12, 12, 160)  179200      activation_47[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_43 (BatchNo (None, 12, 12, 160)  480         conv2d_43[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_48 (BatchNo (None, 12, 12, 160)  480         conv2d_48[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_43 (Activation)      (None, 12, 12, 160)  0           batch_normalization_43[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_48 (Activation)      (None, 12, 12, 160)  0           batch_normalization_48[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_5 (AveragePoo (None, 12, 12, 768)  0           mixed4[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_41 (Conv2D)              (None, 12, 12, 192)  147456      mixed4[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_44 (Conv2D)              (None, 12, 12, 192)  215040      activation_43[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_49 (Conv2D)              (None, 12, 12, 192)  215040      activation_48[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_50 (Conv2D)              (None, 12, 12, 192)  147456      average_pooling2d_5[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_41 (BatchNo (None, 12, 12, 192)  576         conv2d_41[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_44 (BatchNo (None, 12, 12, 192)  576         conv2d_44[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_49 (BatchNo (None, 12, 12, 192)  576         conv2d_49[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_50 (BatchNo (None, 12, 12, 192)  576         conv2d_50[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_41 (Activation)      (None, 12, 12, 192)  0           batch_normalization_41[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_44 (Activation)      (None, 12, 12, 192)  0           batch_normalization_44[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_49 (Activation)      (None, 12, 12, 192)  0           batch_normalization_49[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_50 (Activation)      (None, 12, 12, 192)  0           batch_normalization_50[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed5 (Concatenate)            (None, 12, 12, 768)  0           activation_41[0][0]              \n",
      "                                                                 activation_44[0][0]              \n",
      "                                                                 activation_49[0][0]              \n",
      "                                                                 activation_50[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_55 (Conv2D)              (None, 12, 12, 160)  122880      mixed5[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_55 (BatchNo (None, 12, 12, 160)  480         conv2d_55[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_55 (Activation)      (None, 12, 12, 160)  0           batch_normalization_55[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_56 (Conv2D)              (None, 12, 12, 160)  179200      activation_55[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_56 (BatchNo (None, 12, 12, 160)  480         conv2d_56[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_56 (Activation)      (None, 12, 12, 160)  0           batch_normalization_56[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_52 (Conv2D)              (None, 12, 12, 160)  122880      mixed5[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_57 (Conv2D)              (None, 12, 12, 160)  179200      activation_56[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_52 (BatchNo (None, 12, 12, 160)  480         conv2d_52[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_57 (BatchNo (None, 12, 12, 160)  480         conv2d_57[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_52 (Activation)      (None, 12, 12, 160)  0           batch_normalization_52[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_57 (Activation)      (None, 12, 12, 160)  0           batch_normalization_57[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_53 (Conv2D)              (None, 12, 12, 160)  179200      activation_52[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_58 (Conv2D)              (None, 12, 12, 160)  179200      activation_57[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_53 (BatchNo (None, 12, 12, 160)  480         conv2d_53[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_58 (BatchNo (None, 12, 12, 160)  480         conv2d_58[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_53 (Activation)      (None, 12, 12, 160)  0           batch_normalization_53[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_58 (Activation)      (None, 12, 12, 160)  0           batch_normalization_58[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_6 (AveragePoo (None, 12, 12, 768)  0           mixed5[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_51 (Conv2D)              (None, 12, 12, 192)  147456      mixed5[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_54 (Conv2D)              (None, 12, 12, 192)  215040      activation_53[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_59 (Conv2D)              (None, 12, 12, 192)  215040      activation_58[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_60 (Conv2D)              (None, 12, 12, 192)  147456      average_pooling2d_6[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_51 (BatchNo (None, 12, 12, 192)  576         conv2d_51[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_54 (BatchNo (None, 12, 12, 192)  576         conv2d_54[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_59 (BatchNo (None, 12, 12, 192)  576         conv2d_59[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_60 (BatchNo (None, 12, 12, 192)  576         conv2d_60[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_51 (Activation)      (None, 12, 12, 192)  0           batch_normalization_51[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_54 (Activation)      (None, 12, 12, 192)  0           batch_normalization_54[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_59 (Activation)      (None, 12, 12, 192)  0           batch_normalization_59[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_60 (Activation)      (None, 12, 12, 192)  0           batch_normalization_60[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed6 (Concatenate)            (None, 12, 12, 768)  0           activation_51[0][0]              \n",
      "                                                                 activation_54[0][0]              \n",
      "                                                                 activation_59[0][0]              \n",
      "                                                                 activation_60[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_65 (Conv2D)              (None, 12, 12, 192)  147456      mixed6[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_65 (BatchNo (None, 12, 12, 192)  576         conv2d_65[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_65 (Activation)      (None, 12, 12, 192)  0           batch_normalization_65[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_66 (Conv2D)              (None, 12, 12, 192)  258048      activation_65[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_66 (BatchNo (None, 12, 12, 192)  576         conv2d_66[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_66 (Activation)      (None, 12, 12, 192)  0           batch_normalization_66[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_62 (Conv2D)              (None, 12, 12, 192)  147456      mixed6[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_67 (Conv2D)              (None, 12, 12, 192)  258048      activation_66[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_62 (BatchNo (None, 12, 12, 192)  576         conv2d_62[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_67 (BatchNo (None, 12, 12, 192)  576         conv2d_67[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_62 (Activation)      (None, 12, 12, 192)  0           batch_normalization_62[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_67 (Activation)      (None, 12, 12, 192)  0           batch_normalization_67[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_63 (Conv2D)              (None, 12, 12, 192)  258048      activation_62[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_68 (Conv2D)              (None, 12, 12, 192)  258048      activation_67[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_63 (BatchNo (None, 12, 12, 192)  576         conv2d_63[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_68 (BatchNo (None, 12, 12, 192)  576         conv2d_68[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_63 (Activation)      (None, 12, 12, 192)  0           batch_normalization_63[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_68 (Activation)      (None, 12, 12, 192)  0           batch_normalization_68[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_7 (AveragePoo (None, 12, 12, 768)  0           mixed6[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_61 (Conv2D)              (None, 12, 12, 192)  147456      mixed6[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_64 (Conv2D)              (None, 12, 12, 192)  258048      activation_63[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_69 (Conv2D)              (None, 12, 12, 192)  258048      activation_68[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_70 (Conv2D)              (None, 12, 12, 192)  147456      average_pooling2d_7[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_61 (BatchNo (None, 12, 12, 192)  576         conv2d_61[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_64 (BatchNo (None, 12, 12, 192)  576         conv2d_64[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_69 (BatchNo (None, 12, 12, 192)  576         conv2d_69[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_70 (BatchNo (None, 12, 12, 192)  576         conv2d_70[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_61 (Activation)      (None, 12, 12, 192)  0           batch_normalization_61[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_64 (Activation)      (None, 12, 12, 192)  0           batch_normalization_64[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_69 (Activation)      (None, 12, 12, 192)  0           batch_normalization_69[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_70 (Activation)      (None, 12, 12, 192)  0           batch_normalization_70[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed7 (Concatenate)            (None, 12, 12, 768)  0           activation_61[0][0]              \n",
      "                                                                 activation_64[0][0]              \n",
      "                                                                 activation_69[0][0]              \n",
      "                                                                 activation_70[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_73 (Conv2D)              (None, 12, 12, 192)  147456      mixed7[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_73 (BatchNo (None, 12, 12, 192)  576         conv2d_73[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_73 (Activation)      (None, 12, 12, 192)  0           batch_normalization_73[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_74 (Conv2D)              (None, 12, 12, 192)  258048      activation_73[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_74 (BatchNo (None, 12, 12, 192)  576         conv2d_74[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_74 (Activation)      (None, 12, 12, 192)  0           batch_normalization_74[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_71 (Conv2D)              (None, 12, 12, 192)  147456      mixed7[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_75 (Conv2D)              (None, 12, 12, 192)  258048      activation_74[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_71 (BatchNo (None, 12, 12, 192)  576         conv2d_71[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_75 (BatchNo (None, 12, 12, 192)  576         conv2d_75[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_71 (Activation)      (None, 12, 12, 192)  0           batch_normalization_71[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_75 (Activation)      (None, 12, 12, 192)  0           batch_normalization_75[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_72 (Conv2D)              (None, 5, 5, 320)    552960      activation_71[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_76 (Conv2D)              (None, 5, 5, 192)    331776      activation_75[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_72 (BatchNo (None, 5, 5, 320)    960         conv2d_72[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_76 (BatchNo (None, 5, 5, 192)    576         conv2d_76[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_72 (Activation)      (None, 5, 5, 320)    0           batch_normalization_72[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_76 (Activation)      (None, 5, 5, 192)    0           batch_normalization_76[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2D)  (None, 5, 5, 768)    0           mixed7[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "mixed8 (Concatenate)            (None, 5, 5, 1280)   0           activation_72[0][0]              \n",
      "                                                                 activation_76[0][0]              \n",
      "                                                                 max_pooling2d_4[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_81 (Conv2D)              (None, 5, 5, 448)    573440      mixed8[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_81 (BatchNo (None, 5, 5, 448)    1344        conv2d_81[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_81 (Activation)      (None, 5, 5, 448)    0           batch_normalization_81[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_78 (Conv2D)              (None, 5, 5, 384)    491520      mixed8[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_82 (Conv2D)              (None, 5, 5, 384)    1548288     activation_81[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_78 (BatchNo (None, 5, 5, 384)    1152        conv2d_78[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_82 (BatchNo (None, 5, 5, 384)    1152        conv2d_82[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_78 (Activation)      (None, 5, 5, 384)    0           batch_normalization_78[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_82 (Activation)      (None, 5, 5, 384)    0           batch_normalization_82[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_79 (Conv2D)              (None, 5, 5, 384)    442368      activation_78[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_80 (Conv2D)              (None, 5, 5, 384)    442368      activation_78[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_83 (Conv2D)              (None, 5, 5, 384)    442368      activation_82[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_84 (Conv2D)              (None, 5, 5, 384)    442368      activation_82[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_8 (AveragePoo (None, 5, 5, 1280)   0           mixed8[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_77 (Conv2D)              (None, 5, 5, 320)    409600      mixed8[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_79 (BatchNo (None, 5, 5, 384)    1152        conv2d_79[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_80 (BatchNo (None, 5, 5, 384)    1152        conv2d_80[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_83 (BatchNo (None, 5, 5, 384)    1152        conv2d_83[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_84 (BatchNo (None, 5, 5, 384)    1152        conv2d_84[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_85 (Conv2D)              (None, 5, 5, 192)    245760      average_pooling2d_8[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_77 (BatchNo (None, 5, 5, 320)    960         conv2d_77[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_79 (Activation)      (None, 5, 5, 384)    0           batch_normalization_79[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_80 (Activation)      (None, 5, 5, 384)    0           batch_normalization_80[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_83 (Activation)      (None, 5, 5, 384)    0           batch_normalization_83[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_84 (Activation)      (None, 5, 5, 384)    0           batch_normalization_84[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_85 (BatchNo (None, 5, 5, 192)    576         conv2d_85[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_77 (Activation)      (None, 5, 5, 320)    0           batch_normalization_77[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed9_0 (Concatenate)          (None, 5, 5, 768)    0           activation_79[0][0]              \n",
      "                                                                 activation_80[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_1 (Concatenate)     (None, 5, 5, 768)    0           activation_83[0][0]              \n",
      "                                                                 activation_84[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "activation_85 (Activation)      (None, 5, 5, 192)    0           batch_normalization_85[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed9 (Concatenate)            (None, 5, 5, 2048)   0           activation_77[0][0]              \n",
      "                                                                 mixed9_0[0][0]                   \n",
      "                                                                 concatenate_1[0][0]              \n",
      "                                                                 activation_85[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_90 (Conv2D)              (None, 5, 5, 448)    917504      mixed9[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_90 (BatchNo (None, 5, 5, 448)    1344        conv2d_90[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_90 (Activation)      (None, 5, 5, 448)    0           batch_normalization_90[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_87 (Conv2D)              (None, 5, 5, 384)    786432      mixed9[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_91 (Conv2D)              (None, 5, 5, 384)    1548288     activation_90[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_87 (BatchNo (None, 5, 5, 384)    1152        conv2d_87[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_91 (BatchNo (None, 5, 5, 384)    1152        conv2d_91[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_87 (Activation)      (None, 5, 5, 384)    0           batch_normalization_87[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_91 (Activation)      (None, 5, 5, 384)    0           batch_normalization_91[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_88 (Conv2D)              (None, 5, 5, 384)    442368      activation_87[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_89 (Conv2D)              (None, 5, 5, 384)    442368      activation_87[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_92 (Conv2D)              (None, 5, 5, 384)    442368      activation_91[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_93 (Conv2D)              (None, 5, 5, 384)    442368      activation_91[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "average_pooling2d_9 (AveragePoo (None, 5, 5, 2048)   0           mixed9[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_86 (Conv2D)              (None, 5, 5, 320)    655360      mixed9[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_88 (BatchNo (None, 5, 5, 384)    1152        conv2d_88[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_89 (BatchNo (None, 5, 5, 384)    1152        conv2d_89[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_92 (BatchNo (None, 5, 5, 384)    1152        conv2d_92[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_93 (BatchNo (None, 5, 5, 384)    1152        conv2d_93[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_94 (Conv2D)              (None, 5, 5, 192)    393216      average_pooling2d_9[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_86 (BatchNo (None, 5, 5, 320)    960         conv2d_86[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_88 (Activation)      (None, 5, 5, 384)    0           batch_normalization_88[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_89 (Activation)      (None, 5, 5, 384)    0           batch_normalization_89[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_92 (Activation)      (None, 5, 5, 384)    0           batch_normalization_92[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "activation_93 (Activation)      (None, 5, 5, 384)    0           batch_normalization_93[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_94 (BatchNo (None, 5, 5, 192)    576         conv2d_94[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "activation_86 (Activation)      (None, 5, 5, 320)    0           batch_normalization_86[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed9_1 (Concatenate)          (None, 5, 5, 768)    0           activation_88[0][0]              \n",
      "                                                                 activation_89[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_2 (Concatenate)     (None, 5, 5, 768)    0           activation_92[0][0]              \n",
      "                                                                 activation_93[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "activation_94 (Activation)      (None, 5, 5, 192)    0           batch_normalization_94[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "mixed10 (Concatenate)           (None, 5, 5, 2048)   0           activation_86[0][0]              \n",
      "                                                                 mixed9_1[0][0]                   \n",
      "                                                                 concatenate_2[0][0]              \n",
      "                                                                 activation_94[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_1 (Glo (None, 2048)         0           mixed10[0][0]                    \n",
      "==================================================================================================\n",
      "Total params: 21,802,784\n",
      "Trainable params: 21,768,352\n",
      "Non-trainable params: 34,432\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "if not os.path.exists('./data/triplet/train_naive_inceptionV3_features.npy'):\n",
    "    # Load pre-trained InceptionV3\n",
    "    img_size = (224, 224, 3)\n",
    "    inception = InceptionV3(include_top=False, weights='imagenet', input_shape=img_size, pooling='avg')\n",
    "    inception.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Process:          0\n",
      "================================================================================\n",
      "Process:      20000\n",
      "================================================================================\n",
      "Process:      40000\n",
      "================================================================================\n",
      "Process:      60000\n",
      "================================================================================\n",
      "Process:      80000\n",
      "================================================================================\n",
      "Process:     100000\n",
      "========================================================"
     ]
    }
   ],
   "source": [
    "# Process training images\n",
    "if not os.path.exists('./data/triplet/train_naive_inceptionV3_features.npy'):\n",
    "    train_imgs = np.zeros(shape=(len(train_df), 2048), dtype=np.float32)\n",
    "    img_ids = train_df['image_id'].values\n",
    "    steps = 20000\n",
    "    for i in range(0, len(train_df), steps):\n",
    "        tmp_imgs = []\n",
    "        print('\\nProcess: {:10d}'.format(i))\n",
    "\n",
    "        start = i\n",
    "        end = min(len(train_df), i + steps)\n",
    "        for idx in range(start, end):\n",
    "            if idx % 250 == 0:\n",
    "                print('=', end='')\n",
    "\n",
    "            img_id = img_ids[idx]\n",
    "            path = './data/triplet/train/' + str(img_id) + '.jpg'\n",
    "            img = load_img(path, target_size=img_size[:2])\n",
    "            img = img_to_array(img)\n",
    "            tmp_imgs.append(img)\n",
    "\n",
    "        tmp_imgs = np.array(tmp_imgs, dtype=np.float32) / 255.0\n",
    "        tmp_prediction = inception.predict(tmp_imgs)\n",
    "        train_imgs[start: end, ] = tmp_prediction\n",
    "        _ = gc.collect()\n",
    "        \n",
    "    # Save to disk\n",
    "    np.save('./data/triplet/train_naive_inceptionV3_features.npy', train_imgs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Process:          0\n",
      "================================================================================\n",
      "Process:       4000\n",
      "================================================================================\n",
      "Process:       8000\n",
      "================================================================================\n",
      "Process:      12000\n",
      "================================================================================\n",
      "Process:      16000\n",
      "================================================================================\n",
      "Process:      20000\n",
      "=============================================="
     ]
    }
   ],
   "source": [
    "# Process validation images\n",
    "if not os.path.exists('./data/triplet/validation_naive_inceptionV3_features.npy'):\n",
    "    val_imgs = np.zeros(shape=(len(val_df), 2048), dtype=np.float32)\n",
    "    img_ids = val_df['image_id'].values\n",
    "    steps = 4000\n",
    "    for i in range(0, len(val_df), steps):\n",
    "        tmp_imgs = []\n",
    "        print('\\nProcess: {:10d}'.format(i))\n",
    "\n",
    "        start = i\n",
    "        end = min(len(val_df), i + steps)\n",
    "        for idx in range(start, end):\n",
    "            if idx % 50 == 0:\n",
    "                print('=', end='')\n",
    "\n",
    "            img_id = img_ids[idx]\n",
    "            path = './data/triplet/validation/' + str(img_id) + '.jpg'\n",
    "            img = load_img(path, target_size=img_size[:2])\n",
    "            img = img_to_array(img)\n",
    "            tmp_imgs.append(img)\n",
    "\n",
    "        tmp_imgs = np.array(tmp_imgs, dtype=np.float32) / 255.0\n",
    "        tmp_prediction = inception.predict(tmp_imgs)\n",
    "        val_imgs[start: end, ] = tmp_prediction\n",
    "        _ = gc.collect()\n",
    "\n",
    "    # Save to disk\n",
    "    np.save('./data/triplet/validation_naive_inceptionV3_features.npy', val_imgs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Process:          0\n",
      "================================================================================\n",
      "Process:       4000\n",
      "================================================================================\n",
      "Process:       8000\n",
      "================================================================================\n",
      "Process:      12000\n",
      "================================================================================\n",
      "Process:      16000\n",
      "================================================================================\n",
      "Process:      20000\n",
      "================================================"
     ]
    }
   ],
   "source": [
    "# Process test images\n",
    "if not os.path.exists('./data/triplet/test_naive_inceptionV3_features.npy'):\n",
    "    test_imgs = np.zeros(shape=(len(test_df), 2048), dtype=np.float32)\n",
    "    img_ids = test_df['image_id'].values\n",
    "    steps = 4000\n",
    "    for i in range(0, len(test_df), steps):\n",
    "        tmp_imgs = []\n",
    "        print('\\nProcess: {:10d}'.format(i))\n",
    "\n",
    "        start = i\n",
    "        end = min(len(test_df), i + steps)\n",
    "        for idx in range(start, end):\n",
    "            if idx % 50 == 0:\n",
    "                print('=', end='')\n",
    "\n",
    "            img_id = img_ids[idx]\n",
    "            path = './data/triplet/test/' + str(img_id) + '.jpg'\n",
    "            img = load_img(path, target_size=img_size[:2])\n",
    "            img = img_to_array(img)\n",
    "            tmp_imgs.append(img)\n",
    "\n",
    "        tmp_imgs = np.array(tmp_imgs, dtype=np.float32) / 255.0\n",
    "        tmp_prediction = inception.predict(tmp_imgs)\n",
    "        test_imgs[start: end, ] = tmp_prediction\n",
    "        _ = gc.collect()\n",
    "\n",
    "    # Save to disk\n",
    "    np.save('./data/triplet/test_naive_inceptionV3_features.npy', test_imgs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Load Features and Labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train:\t\t (113783, 2048) (113783, 4)\n",
      "Validation:\t (22255, 2048) (22255, 4)\n",
      "Test:\t\t (22391, 2048) (22391, 4)\n"
     ]
    }
   ],
   "source": [
    "train_feature = np.load('./data/triplet/train_naive_inceptionV3_features.npy')\n",
    "val_feature = np.load('./data/triplet/validation_naive_inceptionV3_features.npy')\n",
    "test_feature = np.load('./data/triplet/test_naive_inceptionV3_features.npy')\n",
    "\n",
    "train_df = pd.read_csv('./data/triplet/train.csv')\n",
    "val_df = pd.read_csv('./data/triplet/validation.csv')\n",
    "test_df = pd.read_csv('./data/triplet/test.csv')\n",
    "\n",
    "print('Train:\\t\\t', train_feature.shape, train_df.shape)\n",
    "print('Validation:\\t', val_feature.shape, val_df.shape)\n",
    "print('Test:\\t\\t', test_feature.shape, test_df.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Helper function\n",
    "def accuracy(true_label, prediction, top=1):\n",
    "    \"\"\" function to calculate the prediction accuracy \"\"\"\n",
    "    prediction = prediction[:, :top]\n",
    "    count = 0\n",
    "    for i in range(len(true_label)):\n",
    "        if true_label[i] in prediction[i]:\n",
    "            count += 1\n",
    "            \n",
    "    return count / len(true_label)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. Implement KNN Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalize features\n",
    "train_norm = np.linalg.norm(train_feature, axis=1, keepdims=True)\n",
    "val_norm = np.linalg.norm(val_feature, axis=1, keepdims=True)\n",
    "test_norm = np.linalg.norm(test_feature, axis=1, keepdims=True)\n",
    "\n",
    "train_norm_feature = train_feature / train_norm\n",
    "val_norm_feature = val_feature / val_norm\n",
    "test_norm_feature = test_feature / test_norm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Merge train and validation features\n",
    "train_val_norm_feature = np.concatenate((train_norm_feature, val_norm_feature), axis=0)\n",
    "train_val_df = pd.concat((train_df, val_df), axis=0)\n",
    "train_val_df = train_val_df.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Implement KNN model\n",
    "if not os.path.exists('./result/knn_naive_inceptionV3_distance.npy'):\n",
    "    knn = NearestNeighbors(n_neighbors=50, algorithm='auto', leaf_size=30, \n",
    "                           metric='minkowski', p=2, n_jobs=-1)\n",
    "    knn.fit(train_val_norm_feature)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists('./result/knn_naive_inceptionV3_distance.npy'):\n",
    "    # Search the first 50 neighbors\n",
    "    distance, neighbor_index = knn.kneighbors(test_norm_feature, return_distance=True)\n",
    "\n",
    "    # Save the results\n",
    "    np.save('./result/knn_naive_inceptionV3_distance.npy', distance)\n",
    "    np.save('./result/knn_naive_inceptionV3_neighbor.npy', neighbor_index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Search Neighbors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_distance = np.load('./result/knn_naive_inceptionV3_distance.npy')\n",
    "knn_neighbor = np.load('./result/knn_naive_inceptionV3_neighbor.npy')\n",
    "\n",
    "# Get the first 50 neighbors\n",
    "predictions = []\n",
    "for neighbors in knn_neighbor:\n",
    "    predictions.append(train_val_df.loc[neighbors]['landmark_id'].values)\n",
    "\n",
    "predictions = np.array(predictions)\n",
    "np.save('./result/knn_naive_inceptionV3_test_prediction.npy', predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Compute Accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Top  1 accuracy:\t 0.368808896431602\n",
      "Top  5 accuracy:\t 0.5222187486043499\n",
      "Top 10 accuracy:\t 0.5868429279621277\n",
      "Top 20 accuracy:\t 0.6491894064579519\n"
     ]
    }
   ],
   "source": [
    "print('Top  1 accuracy:\\t', accuracy(test_df['landmark_id'].values, predictions, top=1))\n",
    "print('Top  5 accuracy:\\t', accuracy(test_df['landmark_id'].values, predictions, top=5))\n",
    "print('Top 10 accuracy:\\t', accuracy(test_df['landmark_id'].values, predictions, top=10))\n",
    "print('Top 20 accuracy:\\t', accuracy(test_df['landmark_id'].values, predictions, top=20))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_acc = []\n",
    "for i in range(1, 51):\n",
    "    tmp_acc = accuracy(test_df['landmark_id'].values, predictions, top=i)\n",
    "    knn_acc.append(tmp_acc)\n",
    "\n",
    "np.save('./result/knn_naive_inceptionV3_accuracy.npy', knn_acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Environment (conda_tensorflow_p36)",
   "language": "python",
   "name": "conda_tensorflow_p36"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}