VGG16_미세조정.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "vgg16 (Model)                (None, 4, 4, 512)         14714688  \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 8192)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 256)               2097408   \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 3)                 771       \n",
      "=================================================================\n",
      "Total params: 16,812,867\n",
      "Trainable params: 16,812,867\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.applications import VGG16\n",
    "\n",
    "conv_base = VGG16(weights='imagenet',\n",
    "                  include_top=False,\n",
    "                  input_shape=(150, 150, 3))\n",
    "\n",
    "from keras import models\n",
    "from keras import layers\n",
    "import os\n",
    "import numpy as np\n",
    "from keras import optimizers\n",
    "\n",
    "model = models.Sequential()\n",
    "model.add(conv_base)\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(256, activation='relu'))\n",
    "model.add(layers.Dense(3, activation='softmax'))\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 1067 images belonging to 3 classes.\n",
      "Found 120 images belonging to 3 classes.\n",
      "Epoch 1/165\n",
      " - 14s - loss: 1.0051 - acc: 0.5220 - val_loss: 0.8138 - val_acc: 0.7417\n",
      "Epoch 2/165\n",
      " - 12s - loss: 0.8596 - acc: 0.6389 - val_loss: 0.7127 - val_acc: 0.7667\n",
      "Epoch 3/165\n",
      " - 13s - loss: 0.7308 - acc: 0.7240 - val_loss: 0.5801 - val_acc: 0.8000\n",
      "Epoch 4/165\n",
      " - 12s - loss: 0.6733 - acc: 0.7482 - val_loss: 0.5780 - val_acc: 0.8167\n",
      "Epoch 5/165\n",
      " - 13s - loss: 0.5719 - acc: 0.8140 - val_loss: 0.5565 - val_acc: 0.7250\n",
      "Epoch 6/165\n",
      " - 13s - loss: 0.5644 - acc: 0.7980 - val_loss: 0.5286 - val_acc: 0.8333\n",
      "Epoch 7/165\n",
      " - 13s - loss: 0.5462 - acc: 0.8002 - val_loss: 0.4988 - val_acc: 0.8167\n",
      "Epoch 8/165\n",
      " - 13s - loss: 0.4695 - acc: 0.8343 - val_loss: 0.4308 - val_acc: 0.8417\n",
      "Epoch 9/165\n",
      " - 13s - loss: 0.4683 - acc: 0.8480 - val_loss: 0.4171 - val_acc: 0.8250\n",
      "Epoch 10/165\n",
      " - 13s - loss: 0.4619 - acc: 0.8178 - val_loss: 0.4138 - val_acc: 0.8250\n",
      "Epoch 11/165\n",
      " - 14s - loss: 0.4541 - acc: 0.8380 - val_loss: 0.4082 - val_acc: 0.8500\n",
      "Epoch 12/165\n",
      " - 13s - loss: 0.3997 - acc: 0.8720 - val_loss: 0.4134 - val_acc: 0.8250\n",
      "Epoch 13/165\n",
      " - 13s - loss: 0.3928 - acc: 0.8543 - val_loss: 0.4421 - val_acc: 0.8250\n",
      "Epoch 14/165\n",
      " - 13s - loss: 0.3795 - acc: 0.8859 - val_loss: 0.3911 - val_acc: 0.8417\n",
      "Epoch 15/165\n",
      " - 13s - loss: 0.3604 - acc: 0.8720 - val_loss: 0.4430 - val_acc: 0.8167\n",
      "Epoch 16/165\n",
      " - 14s - loss: 0.3690 - acc: 0.8580 - val_loss: 0.3718 - val_acc: 0.8500\n",
      "Epoch 17/165\n",
      " - 12s - loss: 0.3387 - acc: 0.8767 - val_loss: 0.3981 - val_acc: 0.8417\n",
      "Epoch 18/165\n",
      " - 13s - loss: 0.3150 - acc: 0.9023 - val_loss: 0.3867 - val_acc: 0.8500\n",
      "Epoch 19/165\n",
      " - 13s - loss: 0.3585 - acc: 0.8860 - val_loss: 0.4015 - val_acc: 0.8417\n",
      "Epoch 20/165\n",
      " - 13s - loss: 0.3345 - acc: 0.8719 - val_loss: 0.3658 - val_acc: 0.8583\n",
      "Epoch 21/165\n",
      " - 13s - loss: 0.3237 - acc: 0.8900 - val_loss: 0.3973 - val_acc: 0.8250\n",
      "Epoch 22/165\n",
      " - 13s - loss: 0.3179 - acc: 0.8959 - val_loss: 0.4102 - val_acc: 0.8417\n",
      "Epoch 23/165\n",
      " - 13s - loss: 0.3222 - acc: 0.8920 - val_loss: 0.3824 - val_acc: 0.8417\n",
      "Epoch 24/165\n",
      " - 13s - loss: 0.3160 - acc: 0.8860 - val_loss: 0.4166 - val_acc: 0.8250\n",
      "Epoch 25/165\n",
      " - 13s - loss: 0.3090 - acc: 0.8960 - val_loss: 0.4326 - val_acc: 0.8417\n",
      "Epoch 26/165\n",
      " - 13s - loss: 0.2767 - acc: 0.9079 - val_loss: 0.3742 - val_acc: 0.8500\n",
      "Epoch 27/165\n",
      " - 12s - loss: 0.3123 - acc: 0.8919 - val_loss: 0.3757 - val_acc: 0.8500\n",
      "Epoch 28/165\n",
      " - 12s - loss: 0.2930 - acc: 0.9020 - val_loss: 0.3535 - val_acc: 0.8583\n",
      "Epoch 29/165\n",
      " - 13s - loss: 0.2669 - acc: 0.9140 - val_loss: 0.3259 - val_acc: 0.8750\n",
      "Epoch 30/165\n",
      " - 12s - loss: 0.2857 - acc: 0.8903 - val_loss: 0.3471 - val_acc: 0.8583\n",
      "Epoch 31/165\n",
      " - 13s - loss: 0.2960 - acc: 0.8980 - val_loss: 0.3397 - val_acc: 0.8583\n",
      "Epoch 32/165\n",
      " - 13s - loss: 0.2514 - acc: 0.9140 - val_loss: 0.3540 - val_acc: 0.8583\n",
      "Epoch 33/165\n",
      " - 13s - loss: 0.3042 - acc: 0.8847 - val_loss: 0.3532 - val_acc: 0.8583\n",
      "Epoch 34/165\n",
      " - 12s - loss: 0.2950 - acc: 0.9123 - val_loss: 0.3678 - val_acc: 0.8583\n",
      "Epoch 35/165\n",
      " - 13s - loss: 0.2566 - acc: 0.9160 - val_loss: 0.2896 - val_acc: 0.9000\n",
      "Epoch 36/165\n",
      " - 13s - loss: 0.2690 - acc: 0.9000 - val_loss: 0.3055 - val_acc: 0.8750\n",
      "Epoch 37/165\n",
      " - 13s - loss: 0.2349 - acc: 0.9163 - val_loss: 0.3115 - val_acc: 0.8750\n",
      "Epoch 38/165\n",
      " - 13s - loss: 0.2257 - acc: 0.9239 - val_loss: 0.3510 - val_acc: 0.8667\n",
      "Epoch 39/165\n",
      " - 13s - loss: 0.2212 - acc: 0.9360 - val_loss: 0.3720 - val_acc: 0.8583\n",
      "Epoch 40/165\n",
      " - 13s - loss: 0.2124 - acc: 0.9140 - val_loss: 0.3407 - val_acc: 0.8750\n",
      "Epoch 41/165\n",
      " - 12s - loss: 0.2294 - acc: 0.9199 - val_loss: 0.2985 - val_acc: 0.8750\n",
      "Epoch 42/165\n",
      " - 13s - loss: 0.2068 - acc: 0.9320 - val_loss: 0.3580 - val_acc: 0.8417\n",
      "Epoch 43/165\n",
      " - 12s - loss: 0.2133 - acc: 0.9359 - val_loss: 0.3407 - val_acc: 0.8583\n",
      "Epoch 44/165\n",
      " - 13s - loss: 0.2081 - acc: 0.9340 - val_loss: 0.3149 - val_acc: 0.8750\n",
      "Epoch 45/165\n",
      " - 13s - loss: 0.2182 - acc: 0.9260 - val_loss: 0.3355 - val_acc: 0.8667\n",
      "Epoch 46/165\n",
      " - 13s - loss: 0.2327 - acc: 0.9359 - val_loss: 0.3591 - val_acc: 0.8500\n",
      "Epoch 47/165\n",
      " - 13s - loss: 0.2288 - acc: 0.9220 - val_loss: 0.3594 - val_acc: 0.8583\n",
      "Epoch 48/165\n",
      " - 13s - loss: 0.1872 - acc: 0.9460 - val_loss: 0.2923 - val_acc: 0.8833\n",
      "Epoch 49/165\n",
      " - 12s - loss: 0.2352 - acc: 0.8983 - val_loss: 0.3232 - val_acc: 0.8667\n",
      "Epoch 50/165\n",
      " - 13s - loss: 0.1960 - acc: 0.9340 - val_loss: 0.3224 - val_acc: 0.8750\n",
      "Epoch 51/165\n",
      " - 12s - loss: 0.1919 - acc: 0.9439 - val_loss: 0.3260 - val_acc: 0.8667\n",
      "Epoch 52/165\n",
      " - 13s - loss: 0.2391 - acc: 0.9240 - val_loss: 0.3757 - val_acc: 0.8417\n",
      "Epoch 53/165\n",
      " - 12s - loss: 0.2016 - acc: 0.9379 - val_loss: 0.3176 - val_acc: 0.8667\n",
      "Epoch 54/165\n",
      " - 13s - loss: 0.2021 - acc: 0.9400 - val_loss: 0.3219 - val_acc: 0.8583\n",
      "Epoch 55/165\n",
      " - 13s - loss: 0.1908 - acc: 0.9460 - val_loss: 0.3814 - val_acc: 0.8667\n",
      "Epoch 56/165\n",
      " - 12s - loss: 0.2092 - acc: 0.9400 - val_loss: 0.2913 - val_acc: 0.9000\n",
      "Epoch 57/165\n",
      " - 13s - loss: 0.2227 - acc: 0.9327 - val_loss: 0.2978 - val_acc: 0.8750\n",
      "Epoch 58/165\n",
      " - 13s - loss: 0.1758 - acc: 0.9520 - val_loss: 0.2532 - val_acc: 0.9083\n",
      "Epoch 59/165\n",
      " - 13s - loss: 0.2211 - acc: 0.9260 - val_loss: 0.3251 - val_acc: 0.8667\n",
      "Epoch 60/165\n",
      " - 13s - loss: 0.1939 - acc: 0.9380 - val_loss: 0.2863 - val_acc: 0.8750\n",
      "Epoch 61/165\n",
      " - 13s - loss: 0.2061 - acc: 0.9399 - val_loss: 0.3087 - val_acc: 0.8667\n",
      "Epoch 62/165\n",
      " - 13s - loss: 0.1866 - acc: 0.9400 - val_loss: 0.3907 - val_acc: 0.8583\n",
      "Epoch 63/165\n",
      " - 13s - loss: 0.1699 - acc: 0.9484 - val_loss: 0.3214 - val_acc: 0.8583\n",
      "Epoch 64/165\n",
      " - 12s - loss: 0.1756 - acc: 0.9479 - val_loss: 0.2806 - val_acc: 0.8833\n",
      "Epoch 65/165\n",
      " - 13s - loss: 0.1833 - acc: 0.9360 - val_loss: 0.3500 - val_acc: 0.8500\n",
      "Epoch 66/165\n",
      " - 13s - loss: 0.1673 - acc: 0.9500 - val_loss: 0.2873 - val_acc: 0.8750\n",
      "Epoch 67/165\n",
      " - 13s - loss: 0.1570 - acc: 0.9524 - val_loss: 0.3157 - val_acc: 0.8583\n",
      "Epoch 68/165\n",
      " - 12s - loss: 0.1897 - acc: 0.9339 - val_loss: 0.3092 - val_acc: 0.8583\n",
      "Epoch 69/165\n",
      " - 12s - loss: 0.1785 - acc: 0.9460 - val_loss: 0.3555 - val_acc: 0.8500\n",
      "Epoch 70/165\n",
      " - 13s - loss: 0.1684 - acc: 0.9479 - val_loss: 0.2792 - val_acc: 0.8833\n",
      "Epoch 71/165\n",
      " - 12s - loss: 0.1463 - acc: 0.9520 - val_loss: 0.2968 - val_acc: 0.8667\n",
      "Epoch 72/165\n",
      " - 13s - loss: 0.1821 - acc: 0.9400 - val_loss: 0.2734 - val_acc: 0.8917\n",
      "Epoch 73/165\n",
      " - 12s - loss: 0.1668 - acc: 0.9540 - val_loss: 0.3291 - val_acc: 0.8667\n",
      "Epoch 74/165\n",
      " - 13s - loss: 0.1777 - acc: 0.9280 - val_loss: 0.2837 - val_acc: 0.8833\n",
      "Epoch 75/165\n",
      " - 13s - loss: 0.1418 - acc: 0.9540 - val_loss: 0.3103 - val_acc: 0.8750\n",
      "Epoch 76/165\n",
      " - 12s - loss: 0.1879 - acc: 0.9308 - val_loss: 0.2912 - val_acc: 0.8750\n",
      "Epoch 77/165\n",
      " - 13s - loss: 0.1654 - acc: 0.9479 - val_loss: 0.2824 - val_acc: 0.8750\n",
      "Epoch 78/165\n",
      " - 13s - loss: 0.1461 - acc: 0.9620 - val_loss: 0.3178 - val_acc: 0.8750\n",
      "Epoch 79/165\n",
      " - 13s - loss: 0.1524 - acc: 0.9540 - val_loss: 0.3020 - val_acc: 0.8750\n",
      "Epoch 80/165\n",
      " - 13s - loss: 0.1779 - acc: 0.9424 - val_loss: 0.2984 - val_acc: 0.8667\n",
      "Epoch 81/165\n",
      " - 13s - loss: 0.1507 - acc: 0.9560 - val_loss: 0.2718 - val_acc: 0.8750\n",
      "Epoch 82/165\n",
      " - 12s - loss: 0.1849 - acc: 0.9419 - val_loss: 0.3134 - val_acc: 0.8667\n",
      "Epoch 83/165\n",
      " - 13s - loss: 0.1407 - acc: 0.9620 - val_loss: 0.3312 - val_acc: 0.8583\n",
      "Epoch 84/165\n",
      " - 13s - loss: 0.1610 - acc: 0.9560 - val_loss: 0.2884 - val_acc: 0.8667\n",
      "Epoch 85/165\n",
      " - 13s - loss: 0.1272 - acc: 0.9740 - val_loss: 0.2809 - val_acc: 0.8667\n",
      "Epoch 86/165\n",
      " - 13s - loss: 0.1564 - acc: 0.9460 - val_loss: 0.2488 - val_acc: 0.8917\n",
      "Epoch 87/165\n",
      " - 13s - loss: 0.1652 - acc: 0.9520 - val_loss: 0.2334 - val_acc: 0.8917\n",
      "Epoch 88/165\n",
      " - 12s - loss: 0.1591 - acc: 0.9488 - val_loss: 0.2952 - val_acc: 0.8750\n",
      "Epoch 89/165\n",
      " - 13s - loss: 0.1565 - acc: 0.9420 - val_loss: 0.2764 - val_acc: 0.8750\n",
      "Epoch 90/165\n",
      " - 12s - loss: 0.1526 - acc: 0.9580 - val_loss: 0.3058 - val_acc: 0.8667\n",
      "Epoch 91/165\n",
      " - 13s - loss: 0.1271 - acc: 0.9580 - val_loss: 0.3391 - val_acc: 0.8583\n",
      "Epoch 92/165\n",
      " - 12s - loss: 0.1416 - acc: 0.9580 - val_loss: 0.3341 - val_acc: 0.8333\n",
      "Epoch 93/165\n",
      " - 13s - loss: 0.1554 - acc: 0.9460 - val_loss: 0.3671 - val_acc: 0.8500\n",
      "Epoch 94/165\n",
      " - 12s - loss: 0.1323 - acc: 0.9720 - val_loss: 0.3400 - val_acc: 0.8667\n",
      "Epoch 95/165\n",
      " - 12s - loss: 0.1412 - acc: 0.9560 - val_loss: 0.3155 - val_acc: 0.8667\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 96/165\n",
      " - 13s - loss: 0.1129 - acc: 0.9684 - val_loss: 0.2944 - val_acc: 0.8750\n",
      "Epoch 97/165\n",
      " - 12s - loss: 0.1318 - acc: 0.9620 - val_loss: 0.3615 - val_acc: 0.8500\n",
      "Epoch 98/165\n",
      " - 13s - loss: 0.1330 - acc: 0.9620 - val_loss: 0.2897 - val_acc: 0.8833\n",
      "Epoch 99/165\n",
      " - 13s - loss: 0.1504 - acc: 0.9520 - val_loss: 0.2976 - val_acc: 0.8833\n",
      "Epoch 100/165\n",
      " - 12s - loss: 0.1372 - acc: 0.9639 - val_loss: 0.3532 - val_acc: 0.8583\n",
      "Epoch 101/165\n",
      " - 13s - loss: 0.1352 - acc: 0.9540 - val_loss: 0.3431 - val_acc: 0.8583\n",
      "Epoch 102/165\n",
      " - 13s - loss: 0.1583 - acc: 0.9479 - val_loss: 0.3244 - val_acc: 0.8667\n",
      "Epoch 103/165\n",
      " - 13s - loss: 0.1279 - acc: 0.9720 - val_loss: 0.3724 - val_acc: 0.8333\n",
      "Epoch 104/165\n",
      " - 13s - loss: 0.1466 - acc: 0.9500 - val_loss: 0.3439 - val_acc: 0.8750\n",
      "Epoch 105/165\n",
      " - 13s - loss: 0.1263 - acc: 0.9660 - val_loss: 0.3147 - val_acc: 0.8583\n",
      "Epoch 106/165\n",
      " - 13s - loss: 0.1273 - acc: 0.9720 - val_loss: 0.2845 - val_acc: 0.8833\n",
      "Epoch 107/165\n",
      " - 12s - loss: 0.1222 - acc: 0.9620 - val_loss: 0.3272 - val_acc: 0.8500\n",
      "Epoch 108/165\n",
      " - 13s - loss: 0.1265 - acc: 0.9540 - val_loss: 0.3535 - val_acc: 0.8417\n",
      "Epoch 109/165\n",
      " - 12s - loss: 0.1307 - acc: 0.9660 - val_loss: 0.2894 - val_acc: 0.8750\n",
      "Epoch 110/165\n",
      " - 12s - loss: 0.1002 - acc: 0.9740 - val_loss: 0.3145 - val_acc: 0.8667\n",
      "Epoch 111/165\n",
      " - 13s - loss: 0.1422 - acc: 0.9600 - val_loss: 0.3839 - val_acc: 0.8500\n",
      "Epoch 112/165\n",
      " - 13s - loss: 0.1345 - acc: 0.9580 - val_loss: 0.3226 - val_acc: 0.8583\n",
      "Epoch 113/165\n",
      " - 13s - loss: 0.1349 - acc: 0.9700 - val_loss: 0.3355 - val_acc: 0.8583\n",
      "Epoch 114/165\n",
      " - 12s - loss: 0.1251 - acc: 0.9580 - val_loss: 0.3824 - val_acc: 0.8500\n",
      "Epoch 115/165\n",
      " - 13s - loss: 0.1285 - acc: 0.9660 - val_loss: 0.3247 - val_acc: 0.8750\n",
      "Epoch 116/165\n",
      " - 12s - loss: 0.1228 - acc: 0.9680 - val_loss: 0.3221 - val_acc: 0.8750\n",
      "Epoch 117/165\n",
      " - 13s - loss: 0.1187 - acc: 0.9640 - val_loss: 0.3029 - val_acc: 0.8750\n",
      "Epoch 118/165\n",
      " - 13s - loss: 0.1295 - acc: 0.9740 - val_loss: 0.3103 - val_acc: 0.8667\n",
      "Epoch 119/165\n",
      " - 13s - loss: 0.1147 - acc: 0.9680 - val_loss: 0.2846 - val_acc: 0.8833\n",
      "Epoch 120/165\n",
      " - 13s - loss: 0.1201 - acc: 0.9620 - val_loss: 0.3705 - val_acc: 0.8750\n",
      "Epoch 121/165\n",
      " - 13s - loss: 0.1315 - acc: 0.9640 - val_loss: 0.2546 - val_acc: 0.8833\n",
      "Epoch 122/165\n",
      " - 13s - loss: 0.1212 - acc: 0.9600 - val_loss: 0.3145 - val_acc: 0.8750\n",
      "Epoch 123/165\n",
      " - 12s - loss: 0.1023 - acc: 0.9740 - val_loss: 0.2930 - val_acc: 0.8833\n",
      "Epoch 124/165\n",
      " - 13s - loss: 0.0857 - acc: 0.9820 - val_loss: 0.2897 - val_acc: 0.8750\n",
      "Epoch 125/165\n",
      " - 12s - loss: 0.1162 - acc: 0.9720 - val_loss: 0.3155 - val_acc: 0.8750\n",
      "Epoch 126/165\n",
      " - 13s - loss: 0.1247 - acc: 0.9700 - val_loss: 0.3591 - val_acc: 0.8750\n",
      "Epoch 127/165\n",
      " - 12s - loss: 0.1105 - acc: 0.9684 - val_loss: 0.2893 - val_acc: 0.8750\n",
      "Epoch 128/165\n",
      " - 13s - loss: 0.0970 - acc: 0.9800 - val_loss: 0.3047 - val_acc: 0.8750\n",
      "Epoch 129/165\n",
      " - 13s - loss: 0.1156 - acc: 0.9680 - val_loss: 0.2920 - val_acc: 0.8750\n",
      "Epoch 130/165\n",
      " - 13s - loss: 0.1410 - acc: 0.9504 - val_loss: 0.2999 - val_acc: 0.8667\n",
      "Epoch 131/165\n",
      " - 12s - loss: 0.1105 - acc: 0.9780 - val_loss: 0.3398 - val_acc: 0.8667\n",
      "Epoch 132/165\n",
      " - 13s - loss: 0.0961 - acc: 0.9820 - val_loss: 0.2646 - val_acc: 0.8750\n",
      "Epoch 133/165\n",
      " - 12s - loss: 0.0881 - acc: 0.9720 - val_loss: 0.3717 - val_acc: 0.8667\n",
      "Epoch 134/165\n",
      " - 13s - loss: 0.0946 - acc: 0.9840 - val_loss: 0.2820 - val_acc: 0.8750\n",
      "Epoch 135/165\n",
      " - 12s - loss: 0.1095 - acc: 0.9700 - val_loss: 0.3346 - val_acc: 0.8750\n",
      "Epoch 136/165\n",
      " - 12s - loss: 0.0951 - acc: 0.9780 - val_loss: 0.3006 - val_acc: 0.8750\n",
      "Epoch 137/165\n",
      " - 13s - loss: 0.0989 - acc: 0.9740 - val_loss: 0.4749 - val_acc: 0.8333\n",
      "Epoch 138/165\n",
      " - 12s - loss: 0.1241 - acc: 0.9620 - val_loss: 0.3085 - val_acc: 0.8750\n",
      "Epoch 139/165\n",
      " - 13s - loss: 0.0988 - acc: 0.9720 - val_loss: 0.2964 - val_acc: 0.8833\n",
      "Epoch 140/165\n",
      " - 13s - loss: 0.1121 - acc: 0.9620 - val_loss: 0.3411 - val_acc: 0.8750\n",
      "Epoch 141/165\n",
      " - 13s - loss: 0.0854 - acc: 0.9904 - val_loss: 0.3230 - val_acc: 0.8667\n",
      "Epoch 142/165\n",
      " - 13s - loss: 0.1286 - acc: 0.9560 - val_loss: 0.3384 - val_acc: 0.8750\n",
      "Epoch 143/165\n",
      " - 12s - loss: 0.1131 - acc: 0.9660 - val_loss: 0.2936 - val_acc: 0.8833\n",
      "Epoch 144/165\n",
      " - 13s - loss: 0.1028 - acc: 0.9740 - val_loss: 0.3690 - val_acc: 0.8583\n",
      "Epoch 145/165\n",
      " - 12s - loss: 0.0878 - acc: 0.9740 - val_loss: 0.3047 - val_acc: 0.8750\n",
      "Epoch 146/165\n",
      " - 13s - loss: 0.1008 - acc: 0.9680 - val_loss: 0.3600 - val_acc: 0.8750\n",
      "Epoch 147/165\n",
      " - 13s - loss: 0.0746 - acc: 0.9920 - val_loss: 0.3284 - val_acc: 0.8750\n",
      "Epoch 148/165\n",
      " - 13s - loss: 0.0854 - acc: 0.9800 - val_loss: 0.3178 - val_acc: 0.8667\n",
      "Epoch 149/165\n",
      " - 12s - loss: 0.1229 - acc: 0.9620 - val_loss: 0.3963 - val_acc: 0.8833\n",
      "Epoch 150/165\n",
      " - 13s - loss: 0.0900 - acc: 0.9740 - val_loss: 0.3535 - val_acc: 0.8667\n",
      "Epoch 151/165\n",
      " - 12s - loss: 0.0930 - acc: 0.9800 - val_loss: 0.3189 - val_acc: 0.8667\n",
      "Epoch 152/165\n",
      " - 13s - loss: 0.0951 - acc: 0.9760 - val_loss: 0.2995 - val_acc: 0.8750\n",
      "Epoch 153/165\n",
      " - 12s - loss: 0.1031 - acc: 0.9744 - val_loss: 0.3758 - val_acc: 0.8583\n",
      "Epoch 154/165\n",
      " - 13s - loss: 0.0847 - acc: 0.9800 - val_loss: 0.3600 - val_acc: 0.8667\n",
      "Epoch 155/165\n",
      " - 12s - loss: 0.0902 - acc: 0.9840 - val_loss: 0.3241 - val_acc: 0.8583\n",
      "Epoch 156/165\n",
      " - 13s - loss: 0.1081 - acc: 0.9700 - val_loss: 0.4036 - val_acc: 0.8667\n",
      "Epoch 157/165\n",
      " - 13s - loss: 0.1023 - acc: 0.9680 - val_loss: 0.3290 - val_acc: 0.8750\n",
      "Epoch 158/165\n",
      " - 13s - loss: 0.1088 - acc: 0.9740 - val_loss: 0.3461 - val_acc: 0.8750\n",
      "Epoch 159/165\n",
      " - 13s - loss: 0.0875 - acc: 0.9800 - val_loss: 0.3734 - val_acc: 0.8750\n",
      "Epoch 160/165\n",
      " - 12s - loss: 0.0847 - acc: 0.9800 - val_loss: 0.2945 - val_acc: 0.8750\n",
      "Epoch 161/165\n",
      " - 13s - loss: 0.0786 - acc: 0.9780 - val_loss: 0.3974 - val_acc: 0.8750\n",
      "Epoch 162/165\n",
      " - 13s - loss: 0.0892 - acc: 0.9780 - val_loss: 0.2842 - val_acc: 0.8917\n",
      "Epoch 163/165\n",
      " - 13s - loss: 0.0941 - acc: 0.9700 - val_loss: 0.3383 - val_acc: 0.8750\n",
      "Epoch 164/165\n",
      " - 12s - loss: 0.1016 - acc: 0.9680 - val_loss: 0.2854 - val_acc: 0.8833\n",
      "Epoch 165/165\n",
      " - 13s - loss: 0.0874 - acc: 0.9699 - val_loss: 0.3505 - val_acc: 0.8750\n"
     ]
    }
   ],
   "source": [
    "\n",
    "base_dir = './dataset7'\n",
    "\n",
    "train_dir = os.path.join(base_dir, 'train')\n",
    "validation_dir = os.path.join(base_dir, 'validation')\n",
    "test_dir = os.path.join(base_dir, 'test')\n",
    "\n",
    "batch_size = 20\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "\n",
    "train_datagen = ImageDataGenerator(\n",
    "      rescale=1./255,\n",
    "      rotation_range=20,\n",
    "      width_shift_range=0.1,\n",
    "      height_shift_range=0.1,\n",
    "      shear_range=0.1,\n",
    "      zoom_range=0.1,\n",
    "      horizontal_flip=True,\n",
    "      fill_mode='nearest')\n",
    "\n",
    "# 검증 데이터는 증식되어서는 안 됩니다!\n",
    "test_datagen = ImageDataGenerator(rescale=1./255)\n",
    "\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "        # 타깃 디렉터리\n",
    "        train_dir,\n",
    "        # 모든 이미지의 크기를 150 × 150로 변경합니다\n",
    "        target_size=(150, 150),\n",
    "        batch_size=20,\n",
    "        # binary_crossentropy 손실을 사용하므로 이진 레이블이 필요합니다\n",
    "        class_mode='categorical')\n",
    "\n",
    "validation_generator = test_datagen.flow_from_directory(\n",
    "        validation_dir,\n",
    "        target_size=(150, 150),\n",
    "        batch_size=20,\n",
    "        class_mode='categorical')\n",
    "\n",
    "#unfreeze_weights(conv_base, from \"block3_conv1\")\n",
    "conv_base.trainable = False\n",
    "\n",
    "set_trainable = False\n",
    "for layer in conv_base.layers:\n",
    "    if layer.name == 'block5_*':\n",
    "        set_trainable = True\n",
    "    if set_trainable :\n",
    "        layer.trainable = True\n",
    "    else :\n",
    "        layer.trainable = False\n",
    "        \n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=optimizers.Adam(lr=2e-5),#optimizers.RMSprop(lr=2e-5),\n",
    "              metrics=['acc'])\n",
    "\n",
    "history = model.fit_generator(\n",
    "      train_generator,\n",
    "      steps_per_epoch=25,\n",
    "      epochs=165,\n",
    "      validation_data=validation_generator,\n",
    "      validation_steps=3,\n",
    "      verbose=2)\n",
    "# 1)\n",
    "model_json = model.to_json()\n",
    "with open(\"model_0216.json\", \"w\") as json_file : \n",
    "    json_file.write(model_json)\n",
    "\n",
    "model.save_weights('vgg16_0426.h5')\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# 2)\n",
    "model.save('Capstone_vgg16_미세조정_0426.h5')\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(len(acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}