03.01.02.image_classification_Feature_Visualization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CNN Feature map Visualization  \n",
    "\n",
    "지난 실습때 훈련시킨 CNN을 통해 Feature의 의미를 보다 시각적으로 이해해 보고자 합니다.\n",
    "\n",
    "(참고)\n",
    "https://github.com/gilbutITbook/006975/blob/master/5.4-visualizing-what-convnets-learn.ipynb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "import json\n",
    "import numpy as np\n",
    "from keras import models\n",
    "from keras.models import model_from_json\n",
    "from keras.datasets import mnist\n",
    "from keras.utils import to_categorical\n",
    "from keras.optimizers import RMSprop\n",
    "\n",
    "import matplotlib.pyplot as plt #for plotting\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "이전 실습에서 저장한 model의 weight를 로딩해 봅니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\uhmpp\\Anaconda3\\envs\\py36\\lib\\site-packages\\tensorflow\\python\\framework\\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From C:\\Users\\uhmpp\\Anaconda3\\envs\\py36\\lib\\site-packages\\keras\\backend\\tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 28, 28, 32)        832       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 28, 28, 32)        25632     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 14, 14, 64)        18496     \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 14, 14, 64)        36928     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 7, 7, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 3136)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 256)               803072    \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 10)                2570      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 10)                0         \n",
      "=================================================================\n",
      "Total params: 887,530\n",
      "Trainable params: 887,530\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "with open('checkpoints/cnn_model.json','r') as f:\n",
    "    model_json = json.load(f)\n",
    "#     print(model_json)\n",
    "\n",
    "model = model_from_json(str(model_json))\n",
    "model.load_weights('checkpoints/cnn_model.h5')\n",
    "model.summary()  # 기억을 되살리기 위해서 모델 구조를 출력합니다\n",
    "# Define the optimizer\n",
    "optimizer = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.0)\n",
    "# Compile the model\n",
    "model.compile(optimizer = optimizer , loss = \"categorical_crossentropy\", metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 2s 211us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.9937"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "input_shape = (x_train.shape[1], x_train.shape[2], 1)\n",
    "x_train_shape = (-1,) + input_shape\n",
    "x_test = x_test.reshape(x_train_shape)\n",
    "x_test = x_test.astype('float32') / 255\n",
    "y_test = to_categorical(y_test, num_classes = 10)\n",
    "test_loss, test_acc = model.evaluate(x_test, y_test)\n",
    "test_acc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "학습한 모델이 feature를 어떻게 추출해 내는지 시각적으로 살펴보겠습니다.  \n",
    "이를 위해, CNN 모델이 0과 1 이미지를 처리할 때의 중간층(activation map)을 layer별로 시각화해서 비교해 보는 기법을 사용해 보고자 합니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 데이터를 다시 로딩합니다.\n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([7, 2, 1, ..., 4, 5, 6], dtype=uint8)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[3, 10, 13, 25, 28]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# label이 0인 이미지를 찾아보자.\n",
    "y_test_0_idx = [idx for idx, val in enumerate(y_test) if val == 0]\n",
    "y_test_0_idx[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[2, 5, 14, 29, 31]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# label이 1인 이미지를 찾아보자.\n",
    "y_test_1_idx = [idx for idx, val in enumerate(y_test) if val == 1]\n",
    "y_test_1_idx[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x23bbf8686d8>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2016x2016 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(28,28))\n",
    "\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.title(y_test[y_test_0_idx[0]], fontsize=100)\n",
    "plt.imshow(x_test[y_test_0_idx[0]])\n",
    "\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.title(y_test[y_test_1_idx[0]], fontsize=100)\n",
    "plt.imshow(x_test[y_test_1_idx[0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2, 28, 28, 1)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 위 두 이미지를 합쳐서 모델에 한번에 입력해 보자.\n",
    "x_test_input = np.stack((x_test[y_test_0_idx[0]], x_test[y_test_1_idx[0]]))\n",
    "x_test_input = np.expand_dims(x_test_input, axis=3)\n",
    "x_test_input.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_test_input = x_test_input.astype('float32') / 255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[1,\n",
       " 21,\n",
       " 34,\n",
       " 37,\n",
       " 51,\n",
       " 56,\n",
       " 63,\n",
       " 68,\n",
       " 69,\n",
       " 75,\n",
       " 81,\n",
       " 88,\n",
       " 95,\n",
       " 108,\n",
       " 114,\n",
       " 118,\n",
       " 119,\n",
       " 121,\n",
       " 156,\n",
       " 169,\n",
       " 192,\n",
       " 206,\n",
       " 209,\n",
       " 210,\n",
       " 216,\n",
       " 229,\n",
       " 232,\n",
       " 234,\n",
       " 246,\n",
       " 249,\n",
       " 260,\n",
       " 283,\n",
       " 293,\n",
       " 296,\n",
       " 303,\n",
       " 320,\n",
       " 326,\n",
       " 359,\n",
       " 399,\n",
       " 427,\n",
       " 429,\n",
       " 435,\n",
       " 440,\n",
       " 451,\n",
       " 453,\n",
       " 458,\n",
       " 462,\n",
       " 464,\n",
       " 473,\n",
       " 489,\n",
       " 519,\n",
       " 524,\n",
       " 526,\n",
       " 527,\n",
       " 542,\n",
       " 577,\n",
       " 582,\n",
       " 596,\n",
       " 603,\n",
       " 612,\n",
       " 633,\n",
       " 639,\n",
       " 656,\n",
       " 662,\n",
       " 666,\n",
       " 667,\n",
       " 668,\n",
       " 669,\n",
       " 689,\n",
       " 702,\n",
       " 709,\n",
       " 712,\n",
       " 733,\n",
       " 743,\n",
       " 745,\n",
       " 776,\n",
       " 781,\n",
       " 787,\n",
       " 790,\n",
       " 818,\n",
       " 825,\n",
       " 849,\n",
       " 859,\n",
       " 860,\n",
       " 869,\n",
       " 872,\n",
       " 889,\n",
       " 903,\n",
       " 927,\n",
       " 943,\n",
       " 949,\n",
       " 952,\n",
       " 957,\n",
       " 965,\n",
       " 979,\n",
       " 984,\n",
       " 997,\n",
       " 1000,\n",
       " 1015,\n",
       " 1018,\n",
       " 1028,\n",
       " 1029,\n",
       " 1040,\n",
       " 1046,\n",
       " 1049,\n",
       " 1076,\n",
       " 1078,\n",
       " 1090,\n",
       " 1093,\n",
       " 1102,\n",
       " 1107,\n",
       " 1128,\n",
       " 1137,\n",
       " 1152,\n",
       " 1168,\n",
       " 1179,\n",
       " 1195,\n",
       " 1209,\n",
       " 1268,\n",
       " 1304,\n",
       " 1310,\n",
       " 1346,\n",
       " 1349,\n",
       " 1359,\n",
       " 1363,\n",
       " 1367,\n",
       " 1368,\n",
       " 1371,\n",
       " 1372,\n",
       " 1377,\n",
       " 1386,\n",
       " 1387,\n",
       " 1403,\n",
       " 1423,\n",
       " 1443,\n",
       " 1454,\n",
       " 1471,\n",
       " 1479,\n",
       " 1489,\n",
       " 1495,\n",
       " 1501,\n",
       " 1502,\n",
       " 1512,\n",
       " 1517,\n",
       " 1530,\n",
       " 1532,\n",
       " 1571,\n",
       " 1578,\n",
       " 1590,\n",
       " 1596,\n",
       " 1600,\n",
       " 1605,\n",
       " 1606,\n",
       " 1625,\n",
       " 1626,\n",
       " 1645,\n",
       " 1664,\n",
       " 1678,\n",
       " 1682,\n",
       " 1701,\n",
       " 1709,\n",
       " 1712,\n",
       " 1723,\n",
       " 1725,\n",
       " 1729,\n",
       " 1742,\n",
       " 1769,\n",
       " 1771,\n",
       " 1775,\n",
       " 1796,\n",
       " 1797,\n",
       " 1798,\n",
       " 1819,\n",
       " 1828,\n",
       " 1837,\n",
       " 1857,\n",
       " 1868,\n",
       " 1876,\n",
       " 1877,\n",
       " 1897,\n",
       " 1904,\n",
       " 1907,\n",
       " 1916,\n",
       " 1926,\n",
       " 1927,\n",
       " 1930,\n",
       " 1956,\n",
       " 1963,\n",
       " 1969,\n",
       " 1995,\n",
       " 1999,\n",
       " 2009,\n",
       " 2051,\n",
       " 2058,\n",
       " 2066,\n",
       " 2079,\n",
       " 2081,\n",
       " 2082,\n",
       " 2084,\n",
       " 2100,\n",
       " 2101,\n",
       " 2112,\n",
       " 2121,\n",
       " 2144,\n",
       " 2147,\n",
       " 2160,\n",
       " 2191,\n",
       " 2192,\n",
       " 2195,\n",
       " 2218,\n",
       " 2245,\n",
       " 2253,\n",
       " 2257,\n",
       " 2269,\n",
       " 2278,\n",
       " 2298,\n",
       " 2310,\n",
       " 2325,\n",
       " 2327,\n",
       " 2333,\n",
       " 2340,\n",
       " 2352,\n",
       " 2353,\n",
       " 2373,\n",
       " 2396,\n",
       " 2403,\n",
       " 2404,\n",
       " 2411,\n",
       " 2435,\n",
       " 2436,\n",
       " 2440,\n",
       " 2473,\n",
       " 2483,\n",
       " 2493,\n",
       " 2500,\n",
       " 2525,\n",
       " 2526,\n",
       " 2528,\n",
       " 2532,\n",
       " 2538,\n",
       " 2539,\n",
       " 2557,\n",
       " 2561,\n",
       " 2581,\n",
       " 2582,\n",
       " 2584,\n",
       " 2586,\n",
       " 2597,\n",
       " 2606,\n",
       " 2615,\n",
       " 2617,\n",
       " 2621,\n",
       " 2624,\n",
       " 2629,\n",
       " 2642,\n",
       " 2709,\n",
       " 2718,\n",
       " 2729,\n",
       " 2736,\n",
       " 2745,\n",
       " 2746,\n",
       " 2765,\n",
       " 2770,\n",
       " 2782,\n",
       " 2806,\n",
       " 2817,\n",
       " 2826,\n",
       " 2839,\n",
       " 2854,\n",
       " 2864,\n",
       " 2873,\n",
       " 2882,\n",
       " 2890,\n",
       " 2897,\n",
       " 2899,\n",
       " 2904,\n",
       " 2914,\n",
       " 2919,\n",
       " 2935,\n",
       " 2944,\n",
       " 2952,\n",
       " 2955,\n",
       " 2974,\n",
       " 2975,\n",
       " 2996,\n",
       " 3001,\n",
       " 3012,\n",
       " 3015,\n",
       " 3016,\n",
       " 3021,\n",
       " 3024,\n",
       " 3035,\n",
       " 3049,\n",
       " 3067,\n",
       " 3106,\n",
       " 3107,\n",
       " 3128,\n",
       " 3131,\n",
       " 3135,\n",
       " 3143,\n",
       " 3160,\n",
       " 3175,\n",
       " 3195,\n",
       " 3198,\n",
       " 3213,\n",
       " 3231,\n",
       " 3241,\n",
       " 3247,\n",
       " 3248,\n",
       " 3259,\n",
       " 3262,\n",
       " 3269,\n",
       " 3286,\n",
       " 3309,\n",
       " 3328,\n",
       " 3337,\n",
       " 3367,\n",
       " 3369,\n",
       " 3376,\n",
       " 3377,\n",
       " 3391,\n",
       " 3396,\n",
       " 3409,\n",
       " 3410,\n",
       " 3429,\n",
       " 3434,\n",
       " 3441,\n",
       " 3443,\n",
       " 3461,\n",
       " 3479,\n",
       " 3490,\n",
       " 3492,\n",
       " 3514,\n",
       " 3516,\n",
       " 3529,\n",
       " 3534,\n",
       " 3541,\n",
       " 3562,\n",
       " 3565,\n",
       " 3568,\n",
       " 3585,\n",
       " 3590,\n",
       " 3603,\n",
       " 3610,\n",
       " 3612,\n",
       " 3622,\n",
       " 3632,\n",
       " 3661,\n",
       " 3667,\n",
       " 3673,\n",
       " 3677,\n",
       " 3691,\n",
       " 3693,\n",
       " 3698,\n",
       " 3702,\n",
       " 3709,\n",
       " 3715,\n",
       " 3720,\n",
       " 3734,\n",
       " 3743,\n",
       " 3776,\n",
       " 3777,\n",
       " 3778,\n",
       " 3806,\n",
       " 3809,\n",
       " 3829,\n",
       " 3849,\n",
       " 3876,\n",
       " 3881,\n",
       " 3882,\n",
       " 3884,\n",
       " 3888,\n",
       " 3896,\n",
       " 3906,\n",
       " 3929,\n",
       " 3937,\n",
       " 3943,\n",
       " 3953,\n",
       " 3964,\n",
       " 4002,\n",
       " 4008,\n",
       " 4009,\n",
       " 4023,\n",
       " 4027,\n",
       " 4044,\n",
       " 4047,\n",
       " 4067,\n",
       " 4082,\n",
       " 4106,\n",
       " 4108,\n",
       " 4115,\n",
       " 4123,\n",
       " 4142,\n",
       " 4145,\n",
       " 4159,\n",
       " 4181,\n",
       " 4201,\n",
       " 4203,\n",
       " 4216,\n",
       " 4218,\n",
       " 4220,\n",
       " 4239,\n",
       " 4245,\n",
       " 4265,\n",
       " 4270,\n",
       " 4279,\n",
       " 4284,\n",
       " 4288,\n",
       " 4289,\n",
       " 4310,\n",
       " 4316,\n",
       " 4321,\n",
       " 4325,\n",
       " 4340,\n",
       " 4343,\n",
       " 4344,\n",
       " 4347,\n",
       " 4356,\n",
       " 4374,\n",
       " 4389,\n",
       " 4415,\n",
       " 4444,\n",
       " 4453,\n",
       " 4455,\n",
       " 4465,\n",
       " 4485,\n",
       " 4488,\n",
       " 4496,\n",
       " 4500,\n",
       " 4505,\n",
       " 4527,\n",
       " 4532,\n",
       " 4535,\n",
       " 4539,\n",
       " 4556,\n",
       " 4563,\n",
       " 4565,\n",
       " 4588,\n",
       " 4597,\n",
       " 4607,\n",
       " 4608,\n",
       " 4624,\n",
       " 4629,\n",
       " 4642,\n",
       " 4653,\n",
       " 4655,\n",
       " 4656,\n",
       " 4682,\n",
       " 4686,\n",
       " 4688,\n",
       " 4713,\n",
       " 4718,\n",
       " 4729,\n",
       " 4738,\n",
       " 4744,\n",
       " 4748,\n",
       " 4749,\n",
       " 4756,\n",
       " 4773,\n",
       " 4776,\n",
       " 4793,\n",
       " 4804,\n",
       " 4849,\n",
       " 4852,\n",
       " 4854,\n",
       " 4855,\n",
       " 4870,\n",
       " 4889,\n",
       " 4892,\n",
       " 4906,\n",
       " 4911,\n",
       " 4918,\n",
       " 4926,\n",
       " 4931,\n",
       " 4951,\n",
       " 4962,\n",
       " 4981,\n",
       " 4985,\n",
       " 4994,\n",
       " 5007,\n",
       " 5010,\n",
       " 5015,\n",
       " 5019,\n",
       " 5048,\n",
       " 5052,\n",
       " 5053,\n",
       " 5072,\n",
       " 5074,\n",
       " 5082,\n",
       " 5083,\n",
       " 5096,\n",
       " 5108,\n",
       " 5115,\n",
       " 5120,\n",
       " 5131,\n",
       " 5133,\n",
       " 5144,\n",
       " 5147,\n",
       " 5154,\n",
       " 5167,\n",
       " 5187,\n",
       " 5192,\n",
       " 5194,\n",
       " 5196,\n",
       " 5202,\n",
       " 5203,\n",
       " 5228,\n",
       " 5244,\n",
       " 5246,\n",
       " 5249,\n",
       " 5266,\n",
       " 5268,\n",
       " 5272,\n",
       " 5275,\n",
       " 5285,\n",
       " 5286,\n",
       " 5288,\n",
       " 5298,\n",
       " 5317,\n",
       " 5319,\n",
       " 5321,\n",
       " 5327,\n",
       " 5330,\n",
       " 5340,\n",
       " 5342,\n",
       " 5344,\n",
       " 5353,\n",
       " 5368,\n",
       " 5373,\n",
       " 5393,\n",
       " 5398,\n",
       " 5461,\n",
       " 5462,\n",
       " 5464,\n",
       " 5468,\n",
       " 5469,\n",
       " 5470,\n",
       " 5475,\n",
       " 5487,\n",
       " 5488,\n",
       " 5498,\n",
       " 5502,\n",
       " 5505,\n",
       " 5512,\n",
       " 5514,\n",
       " 5525,\n",
       " 5541,\n",
       " 5560,\n",
       " 5563,\n",
       " 5583,\n",
       " 5585,\n",
       " 5604,\n",
       " 5608,\n",
       " 5612,\n",
       " 5615,\n",
       " 5622,\n",
       " 5627,\n",
       " 5641,\n",
       " 5646,\n",
       " 5662,\n",
       " 5665,\n",
       " 5674,\n",
       " 5690,\n",
       " 5692,\n",
       " 5694,\n",
       " 5697,\n",
       " 5698,\n",
       " 5711,\n",
       " 5729,\n",
       " 5730,\n",
       " 5747,\n",
       " 5767,\n",
       " 5773,\n",
       " 5801,\n",
       " 5808,\n",
       " 5809,\n",
       " 5830,\n",
       " 5844,\n",
       " 5881,\n",
       " 5883,\n",
       " 5884,\n",
       " 5886,\n",
       " 5888,\n",
       " 5905,\n",
       " 5921,\n",
       " 5927,\n",
       " 5929,\n",
       " 5932,\n",
       " 5938,\n",
       " 5961,\n",
       " 5978,\n",
       " 5981,\n",
       " 6001,\n",
       " 6008,\n",
       " 6023,\n",
       " 6034,\n",
       " 6045,\n",
       " 6065,\n",
       " 6076,\n",
       " 6094,\n",
       " 6097,\n",
       " 6103,\n",
       " 6113,\n",
       " 6117,\n",
       " 6133,\n",
       " 6137,\n",
       " 6146,\n",
       " 6179,\n",
       " 6207,\n",
       " 6227,\n",
       " 6228,\n",
       " 6245,\n",
       " 6279,\n",
       " 6292,\n",
       " 6296,\n",
       " 6302,\n",
       " 6316,\n",
       " 6320,\n",
       " 6325,\n",
       " 6326,\n",
       " 6330,\n",
       " 6331,\n",
       " 6332,\n",
       " 6337,\n",
       " 6341,\n",
       " 6357,\n",
       " 6367,\n",
       " 6394,\n",
       " 6400,\n",
       " 6406,\n",
       " 6409,\n",
       " 6421,\n",
       " 6422,\n",
       " 6441,\n",
       " 6444,\n",
       " 6446,\n",
       " 6461,\n",
       " 6468,\n",
       " 6496,\n",
       " 6502,\n",
       " 6538,\n",
       " 6591,\n",
       " 6593,\n",
       " 6603,\n",
       " 6607,\n",
       " 6611,\n",
       " 6614,\n",
       " 6615,\n",
       " 6619,\n",
       " 6629,\n",
       " 6630,\n",
       " 6645,\n",
       " 6648,\n",
       " 6663,\n",
       " 6676,\n",
       " 6681,\n",
       " 6684,\n",
       " 6700,\n",
       " 6704,\n",
       " 6708,\n",
       " 6723,\n",
       " 6730,\n",
       " 6737,\n",
       " 6738,\n",
       " 6757,\n",
       " 6758,\n",
       " 6763,\n",
       " 6768,\n",
       " 6791,\n",
       " 6802,\n",
       " 6819,\n",
       " 6823,\n",
       " 6832,\n",
       " 6847,\n",
       " 6857,\n",
       " 6875,\n",
       " 6883,\n",
       " 6917,\n",
       " 6923,\n",
       " 6928,\n",
       " 6935,\n",
       " 6958,\n",
       " 6960,\n",
       " 6967,\n",
       " 6968,\n",
       " 6970,\n",
       " 6972,\n",
       " 6973,\n",
       " 6979,\n",
       " 6987,\n",
       " 6991,\n",
       " 7003,\n",
       " 7004,\n",
       " 7012,\n",
       " 7018,\n",
       " 7025,\n",
       " 7037,\n",
       " 7038,\n",
       " 7050,\n",
       " 7052,\n",
       " 7055,\n",
       " 7062,\n",
       " 7066,\n",
       " 7127,\n",
       " 7128,\n",
       " 7129,\n",
       " 7130,\n",
       " 7135,\n",
       " 7143,\n",
       " 7158,\n",
       " 7164,\n",
       " 7181,\n",
       " 7183,\n",
       " 7203,\n",
       " 7219,\n",
       " 7243,\n",
       " 7247,\n",
       " 7254,\n",
       " 7267,\n",
       " 7271,\n",
       " 7272,\n",
       " 7281,\n",
       " 7286,\n",
       " 7305,\n",
       " 7318,\n",
       " 7325,\n",
       " 7338,\n",
       " 7355,\n",
       " 7358,\n",
       " 7364,\n",
       " 7371,\n",
       " 7372,\n",
       " 7373,\n",
       " 7378,\n",
       " 7391,\n",
       " 7409,\n",
       " 7427,\n",
       " 7451,\n",
       " 7458,\n",
       " 7462,\n",
       " 7465,\n",
       " 7466,\n",
       " 7471,\n",
       " 7485,\n",
       " 7499,\n",
       " 7506,\n",
       " 7515,\n",
       " 7548,\n",
       " 7552,\n",
       " 7557,\n",
       " 7566,\n",
       " 7585,\n",
       " 7589,\n",
       " 7597,\n",
       " 7629,\n",
       " 7648,\n",
       " 7649,\n",
       " 7664,\n",
       " 7669,\n",
       " 7685,\n",
       " 7687,\n",
       " 7691,\n",
       " 7707,\n",
       " 7708,\n",
       " 7709,\n",
       " 7710,\n",
       " 7728,\n",
       " 7753,\n",
       " 7786,\n",
       " 7807,\n",
       " 7813,\n",
       " 7816,\n",
       " 7830,\n",
       " 7835,\n",
       " 7846,\n",
       " 7855,\n",
       " 7858,\n",
       " 7874,\n",
       " 7875,\n",
       " 7882,\n",
       " 7892,\n",
       " 7894,\n",
       " 7905,\n",
       " 7911,\n",
       " 7925,\n",
       " 7930,\n",
       " 7932,\n",
       " 7934,\n",
       " 7944,\n",
       " 7948,\n",
       " 7954,\n",
       " 7969,\n",
       " 7978,\n",
       " 7981,\n",
       " 7985,\n",
       " 8000,\n",
       " 8003,\n",
       " 8016,\n",
       " 8019,\n",
       " 8021,\n",
       " 8028,\n",
       " 8046,\n",
       " 8050,\n",
       " 8088,\n",
       " 8089,\n",
       " 8093,\n",
       " 8096,\n",
       " 8098,\n",
       " 8108,\n",
       " 8110,\n",
       " 8113,\n",
       " 8116,\n",
       " 8129,\n",
       " 8134,\n",
       " 8147,\n",
       " 8157,\n",
       " 8158,\n",
       " 8159,\n",
       " 8172,\n",
       " 8187,\n",
       " 8193,\n",
       " 8223,\n",
       " 8233,\n",
       " 8247,\n",
       " 8249,\n",
       " 8250,\n",
       " 8253,\n",
       " 8262,\n",
       " 8272,\n",
       " 8273,\n",
       " 8274,\n",
       " 8284,\n",
       " 8288,\n",
       " 8289,\n",
       " 8294,\n",
       " 8304,\n",
       " 8309,\n",
       " 8323,\n",
       " 8335,\n",
       " 8343,\n",
       " 8349,\n",
       " 8354,\n",
       " 8355,\n",
       " 8368,\n",
       " 8369,\n",
       " 8385,\n",
       " 8387,\n",
       " 8449,\n",
       " 8469,\n",
       " 8472,\n",
       " 8489,\n",
       " 8509,\n",
       " 8515,\n",
       " 8527,\n",
       " 8531,\n",
       " 8533,\n",
       " 8540,\n",
       " 8545,\n",
       " 8612,\n",
       " 8653,\n",
       " 8681,\n",
       " 8682,\n",
       " 8687,\n",
       " 8703,\n",
       " 8723,\n",
       " 8724,\n",
       " 8739,\n",
       " 8754,\n",
       " 8763,\n",
       " 8788,\n",
       " 8809,\n",
       " 8812,\n",
       " 8813,\n",
       " 8838,\n",
       " 8855,\n",
       " 8863,\n",
       " 8877,\n",
       " 8885,\n",
       " 8894,\n",
       " 8896,\n",
       " 8897,\n",
       " 8901,\n",
       " 8912,\n",
       " 8917,\n",
       " 8926,\n",
       " 8932,\n",
       " 8935,\n",
       " 8961,\n",
       " 8969,\n",
       " 8979,\n",
       " 8982,\n",
       " 9003,\n",
       " 9005,\n",
       " 9018,\n",
       " 9027,\n",
       " 9032,\n",
       " 9034,\n",
       " 9035,\n",
       " 9038,\n",
       " 9047,\n",
       " 9048,\n",
       " 9055,\n",
       " 9056,\n",
       " 9067,\n",
       " 9083,\n",
       " 9101,\n",
       " 9131,\n",
       " 9142,\n",
       " 9151,\n",
       " 9158,\n",
       " 9166,\n",
       " 9194,\n",
       " 9195,\n",
       " 9196,\n",
       " 9197,\n",
       " 9204,\n",
       " 9209,\n",
       " 9215,\n",
       " 9222,\n",
       " 9257,\n",
       " 9269,\n",
       " 9287,\n",
       " 9296,\n",
       " 9297,\n",
       " 9310,\n",
       " 9313,\n",
       " 9322,\n",
       " 9329,\n",
       " 9342,\n",
       " 9402,\n",
       " 9406,\n",
       " 9422,\n",
       " 9435,\n",
       " 9439,\n",
       " 9440,\n",
       " 9441,\n",
       " 9451,\n",
       " 9452,\n",
       " 9455,\n",
       " 9456,\n",
       " 9458,\n",
       " 9459,\n",
       " 9463,\n",
       " 9473,\n",
       " 9476,\n",
       " 9478,\n",
       " 9485,\n",
       " 9505,\n",
       " 9506,\n",
       " 9523,\n",
       " 9524,\n",
       " 9530,\n",
       " 9544,\n",
       " 9552,\n",
       " 9666,\n",
       " 9671,\n",
       " 9675,\n",
       " 9677,\n",
       " 9687,\n",
       " 9688,\n",
       " 9691,\n",
       " 9695,\n",
       " 9698,\n",
       " 9699,\n",
       " 9703,\n",
       " 9707,\n",
       " 9710,\n",
       " 9717,\n",
       " 9720,\n",
       " 9733,\n",
       " 9742,\n",
       " 9744,\n",
       " 9750,\n",
       " 9753,\n",
       " 9773,\n",
       " 9790,\n",
       " 9806,\n",
       " 9820,\n",
       " 9822,\n",
       " 9823,\n",
       " 9834,\n",
       " 9840,\n",
       " 9857,\n",
       " 9865,\n",
       " 9869,\n",
       " 9874,\n",
       " 9877,\n",
       " 9881,\n",
       " 9890,\n",
       " 9914,\n",
       " 9922,\n",
       " 9934,\n",
       " 9963,\n",
       " 9965,\n",
       " 9966,\n",
       " 9971,\n",
       " 9978,\n",
       " 9983,\n",
       " 9985,\n",
       " 9986,\n",
       " ...]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_0_idx = [idx for idx, val in enumerate(y_train) if val == 0]\n",
    "y_train_1_idx = [idx for idx, val in enumerate(y_train) if val == 1]\n",
    "y_train_0_idx+y_train_1_idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 0, 0, ..., 1, 1, 1], dtype=uint8)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train[y_train_0_idx+y_train_1_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<keras.layers.convolutional.Conv2D at 0x23bbf5304e0>,\n",
       " <keras.layers.convolutional.Conv2D at 0x23bbf530b70>,\n",
       " <keras.layers.pooling.MaxPooling2D at 0x23bbf5598d0>,\n",
       " <keras.layers.core.Dropout at 0x23bbf530a90>,\n",
       " <keras.layers.convolutional.Conv2D at 0x23bbf598da0>,\n",
       " <keras.layers.convolutional.Conv2D at 0x23bbf559b38>,\n",
       " <keras.layers.pooling.MaxPooling2D at 0x23bbf5f9978>,\n",
       " <keras.layers.core.Dropout at 0x23bbf5afa90>,\n",
       " <keras.layers.core.Flatten at 0x23bbf62db70>,\n",
       " <keras.layers.core.Dense at 0x23bbf62d9b0>,\n",
       " <keras.layers.core.Dropout at 0x23bbf66fb00>,\n",
       " <keras.layers.core.Dense at 0x23bbf66fa58>,\n",
       " <keras.layers.core.Activation at 0x23bbf695dd8>]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Tensor 'conv2d_1/Relu:0' shape=(?, 28, 28, 32) dtype=float32>,\n",
       " <tf.Tensor 'conv2d_2/Relu:0' shape=(?, 28, 28, 32) dtype=float32>,\n",
       " <tf.Tensor 'max_pooling2d_1/MaxPool:0' shape=(?, 14, 14, 32) dtype=float32>,\n",
       " <tf.Tensor 'dropout_1/cond/Merge:0' shape=(?, 14, 14, 32) dtype=float32>,\n",
       " <tf.Tensor 'conv2d_3/Relu:0' shape=(?, 14, 14, 64) dtype=float32>,\n",
       " <tf.Tensor 'conv2d_4/Relu:0' shape=(?, 14, 14, 64) dtype=float32>,\n",
       " <tf.Tensor 'max_pooling2d_2/MaxPool:0' shape=(?, 7, 7, 64) dtype=float32>,\n",
       " <tf.Tensor 'dropout_2/cond/Merge:0' shape=(?, 7, 7, 64) dtype=float32>,\n",
       " <tf.Tensor 'flatten_1/Reshape:0' shape=(?, ?) dtype=float32>,\n",
       " <tf.Tensor 'dense_1/Relu:0' shape=(?, 256) dtype=float32>,\n",
       " <tf.Tensor 'dropout_3/cond/Merge:0' shape=(?, 256) dtype=float32>,\n",
       " <tf.Tensor 'dense_2/BiasAdd:0' shape=(?, 10) dtype=float32>,\n",
       " <tf.Tensor 'activation_1/Softmax:0' shape=(?, 10) dtype=float32>]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# layer의 output으로만 이루어진 list를 만들어 봅시다.\n",
    "layer_outputs = [layer.output for layer in model.layers]\n",
    "layer_outputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Tensor 'conv2d_1_input:0' shape=(?, 28, 28, 1) dtype=float32>,\n",
       " <tf.Tensor 'conv2d_1/Relu:0' shape=(?, 28, 28, 32) dtype=float32>,\n",
       " <tf.Tensor 'conv2d_2/Relu:0' shape=(?, 28, 28, 32) dtype=float32>,\n",
       " <tf.Tensor 'max_pooling2d_1/MaxPool:0' shape=(?, 14, 14, 32) dtype=float32>,\n",
       " <tf.Tensor 'dropout_1/cond/Merge:0' shape=(?, 14, 14, 32) dtype=float32>,\n",
       " <tf.Tensor 'conv2d_3/Relu:0' shape=(?, 14, 14, 64) dtype=float32>,\n",
       " <tf.Tensor 'conv2d_4/Relu:0' shape=(?, 14, 14, 64) dtype=float32>,\n",
       " <tf.Tensor 'max_pooling2d_2/MaxPool:0' shape=(?, 7, 7, 64) dtype=float32>,\n",
       " <tf.Tensor 'dropout_2/cond/Merge:0' shape=(?, 7, 7, 64) dtype=float32>,\n",
       " <tf.Tensor 'flatten_1/Reshape:0' shape=(?, ?) dtype=float32>,\n",
       " <tf.Tensor 'dense_1/Relu:0' shape=(?, 256) dtype=float32>,\n",
       " <tf.Tensor 'dropout_3/cond/Merge:0' shape=(?, 256) dtype=float32>,\n",
       " <tf.Tensor 'dense_2/BiasAdd:0' shape=(?, 10) dtype=float32>]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "layer_inputs = [layer.input for layer in model.layers]\n",
    "layer_inputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 입력에 대해 CNN의 각 레이어의 output을 전부 반환하는 모델을 만듭니다:\n",
    "activation_model = models.Model(inputs=model.input, outputs=layer_outputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 층의 활성화마다 하나씩 8개의 넘파이 배열로 이루어진 리스트를 반환합니다:\n",
    "activations  = activation_model.predict(x_test_input)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(activations)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2, 28, 28, 32)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "activations[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 288x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 0 이미지에 대한 1번째 conv 레이어가 출력한 32번째 feature map을 출력해 봅시다.\n",
    "plt.matshow(activations[0][0, :, :, 31], cmap='viridis')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 288x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 1 이미지에 대한 1번째 conv 레이어가 출력한 32번째 feature map을 출력해 봅시다.\n",
    "plt.matshow(activations[0][1, :, :, 31], cmap='viridis')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "def draw_activation_grid(layer_activation, layer_name, images_per_row = 16):\n",
    "    # 특성 맵에 있는 특성의 수\n",
    "    n_features = layer_activation.shape[-1]\n",
    "\n",
    "    # 특성 맵의 크기는 (1, size, size, n_features)입니다\n",
    "    size = layer_activation.shape[1]\n",
    "\n",
    "    # 활성화 채널을 위한 그리드 크기를 구합니다\n",
    "    n_cols = n_features // images_per_row\n",
    "    display_grid = np.zeros((size * n_cols, images_per_row * size))\n",
    "\n",
    "    # 각 활성화를 하나의 큰 그리드에 채웁니다\n",
    "    for col in range(n_cols):\n",
    "        for row in range(images_per_row):\n",
    "            channel_image = layer_activation[\n",
    "                                             :, :,\n",
    "                                             col * images_per_row + row]\n",
    "            # 그래프로 나타내기 좋게 특성을 처리합니다\n",
    "            channel_image -= channel_image.mean()\n",
    "            channel_image /= channel_image.std()\n",
    "            channel_image *= 64\n",
    "            channel_image += 128\n",
    "            channel_image = np.clip(channel_image, 0, 255).astype('uint8')\n",
    "            display_grid[col * size : (col + 1) * size,\n",
    "                         row * size : (row + 1) * size] = channel_image\n",
    "\n",
    "    # 그리드를 출력합니다\n",
    "    scale = 1. / size\n",
    "    plt.figure(figsize=(scale * display_grid.shape[1],\n",
    "                        scale * display_grid.shape[0]))\n",
    "    plt.title(layer_name)\n",
    "    plt.grid(False)\n",
    "    plt.imshow(display_grid, aspect='auto', cmap='viridis')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(56, 448)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "layer_activation = activations[0][0]\n",
    "n_features = layer_activation.shape[-1]\n",
    "n_features\n",
    "size = layer_activation.shape[1]\n",
    "size\n",
    "n_cols = n_features // 16\n",
    "n_cols\n",
    "display_grid = np.zeros((size * n_cols, 16 * size))\n",
    "display_grid.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_activation_grid(activations[1][0], layer_name='Conv2D_layer1_NUMBER:0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\uhmpp\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel_launcher.py:20: RuntimeWarning: invalid value encountered in true_divide\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x144 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_activation_grid(activations[1][1], layer_name='Conv2D_layer1_NUMBER:1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_activation_grid(activations[4][0], layer_name='Conv2D_layer3_NUMBER:0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_activation_grid(activations[4][1], layer_name='Conv2D_layer3_NUMBER:0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\uhmpp\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel_launcher.py:20: RuntimeWarning: invalid value encountered in true_divide\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_activation_grid(activations[7][0], layer_name='Conv2D_layer3_NUMBER:0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\uhmpp\\Anaconda3\\envs\\py36\\lib\\site-packages\\ipykernel_launcher.py:20: RuntimeWarning: invalid value encountered in true_divide\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_activation_grid(activations[7][1], layer_name='Conv2D_layer3_NUMBER:0')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x23d63167fd0>"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x230.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# softmax에 input으로 주어지는 logit vector\n",
    "plt.matshow(activations[11], cmap='viridis')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x23d63226128>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x230.4 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# softmax의 output\n",
    "plt.matshow(activations[12], cmap='viridis')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# softmax의 output\n",
    "plt.matshow(activations[12], cmap='viridis')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "py36",
   "language": "python",
   "name": "py36"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}