8.5-introduction-to-gans.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'2.2.4'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import keras\n",
    "keras.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 적대적 생성 신경망 소개\n",
    "\n",
    "이 노트북은 [케라스 창시자에게 배우는 딥러닝](https://tensorflow.blog/deep-learning-with-python/) 책의 8장 5절의 코드 예제입니다. 책에는 더 많은 내용과 그림이 있습니다. 이 노트북에는 소스 코드에 관련된 설명만 포함합니다. 이 노트북의 설명은 케라스 버전 2.2.2에 맞추어져 있습니다. 케라스 최신 버전이 릴리스되면 노트북을 다시 테스트하기 때문에 설명과 코드의 결과가 조금 다를 수 있습니다.\n",
    "\n",
    "---\n",
    "[...]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## GAN 구현 방법\n",
    "\n",
    "이 절에서는 케라스에서 가장 기본적인 형태의 GAN을 구현하는 방법을 설명하겠습니다. GAN은 수준 높은 기술이기 때문에 기술적인 내용을 깊이 설명하는 것은 이 책의 범위를 벗어납니다. 구체적인 구현은 심층 합성곱 GAN(DCGAN)입니다. 생성자와 판별자가 심층 컨브넷입니다. 특히 생성자에서 이미지 업샘플링을 위해 `Conv2DTranspose` 층을 사용합니다.\n",
    "\n",
    "CIFAR10 데이터셋의 이미지로 GAN을 훈련하겠습니다. 이 데이터셋은 32 × 32 크기의 RGB 이미지 50,000개로 이루어져 있고 10개의 클래스를 가집니다(클래스마다 5,000개의 이미지가 있습니다). 문제를 간단하게 만들기 위해 “frog” 클래스의 이미지만 사용하겠습니다.\n",
    "\n",
    "GAN 구조는 다음과 같습니다:\n",
    "\n",
    "1.\t`generator` 네트워크는 `(latent_dim,)` 크기의 벡터를 `(32, 32, 3)` 크기의 이미지로 매핑합니다.\n",
    "2.\t`discriminator` 네트워크는 (32, 32, 3) 크기의 이미지가 진짜일 확률을 추정하여 이진 값으로 매핑합니다.\n",
    "3.\t생성자와 판별자를 연결하는 `gan` 네트워크를 만듭니다. `gan(x) = discriminator(generator(x))` 입니다. 이 `gan` 네트워크는 잠재 공간의 벡터를 판별자의 평가로 매핑합니다. 판별자는 생성자가 잠재 공간의 벡터를 디코딩한 것이 얼마나 현실적인지를 평가합니다.\n",
    "4.\t“진짜\"/”가짜\" 레이블과 함께 진짜 이미지와 가짜 이미지 샘플을 사용해 판별자를 훈련합니다. 일반적인 이미지 분류 모델을 훈련하는 것과 동일합니다.\n",
    "5. 생성자를 훈련하려면 `gan` 모델의 손실에 대한 생성자 가중치의 그래디언트를 사용합니다. 이 말은 매 단계마다 생성자에 의해 디코딩된 이미지를 판별자가 “진짜\"로 분류하도록 만드는 방향으로 생성자의 가중치를 이동한다는 뜻입니다. 다른 말로하면 판별자를 속이도록 생성자를 훈련합니다."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 훈련 방법\n",
    "\n",
    "GAN을 훈련하고 튜닝하는 과정은 어렵기로 유명합니다. 알아두어야 할 몇 가지 유용한 기법이 있습니다. 딥러닝의 대부분이 그렇듯이 이는 과학보다는 연금술에 가깝습니다. 이런 기법들은 이론에 바탕을 둔 지침이 아니고 경험을 통해 발견된 것입니다. 실제 일어난 현상을 직관적으로 이해하는 수준에서 검증되었습니다. 모든 문제에 반드시 적용해야 것은 아니지만 경험상 잘 작동한다고 알려져 있습니다.\n",
    "\n",
    "다음은 이 절에서 GAN 생성자와 판별자를 구현하는 데 사용할 몇 가지 기법입니다. 이 목록이 GAN에 관련된 전체 팁이 아닙니다. GAN 논문들에서 더 많은 방법을 볼 수 있습니다.\n",
    "\n",
    "* 생성자의 마지막 활성화로 다른 종류의 모델에서 널리 사용하는 `sigmoid` 대신 `tanh` 함수를 사용합니다. \n",
    "* 균등 분포가 아니고 정규 분포(가우시안 분포)를 사용하여 잠재 공간에서 포인트를 샘플링합니다.\n",
    "* 무작위성은 모델을 견고하게 만듭니다. GAN 훈련은 동적 평형을 만들기 때문에 여러 방식으로 갇힐 가능성이 높습니다. 훈련하는 동안 무작위성을 주입하면 이를 방지하는 데 도움이 됩니다. 무작위성은 두 가지 방법으로 주입합니다. 판별자에 드롭아웃을 사용하거나 판별자를 위해 레이블에 랜덤 노이즈를 추가합니다.\n",
    "* 희소한 그래디언트는 GAN 훈련을 방해할 수 있습니다. 딥러닝에서 희소는 종종 바람직한 현상이지만 GAN에서는 그렇지 않습니다. 그래디언트를 희소하게 만들 수 있는 것은 최대 풀링 연산과 ReLU 활성화 두 가지입니다. 최대 풀링 대신 스트라이드 합성곱을 사용해 다운샘플링을 하는 것이 좋습니다. 또 ReLU 활성화 대신 `LeakyReLU` 층을 사용하세요. ReLU와 비슷하지만 음수의 활성화 값을 조금 허용하기 때문에 희소가 조금 완화됩니다.\n",
    "* 생성자에서 픽셀 공간을 균일하게 다루지 못해 생성된 이미지에서 체스판 모양이 종종 나타납니다(그림 8-17 참조). 이를 해결하기 위해 생성자와 판별자에서 스트라이드 `Conv2DTranpose`나 `Conv2D`를 사용할 때 스트라이드 크기로 나누어질 수 있는 커널 크기를 사용합니다."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 생성자\n",
    "\n",
    "먼저 벡터(훈련하는 동안 잠재 공간에서 무작위로 샘플링됩니다)를 후보 이미지로 변환하는 `generator` 모델을 만들어 보죠. GAN에서 발생하는 많은 문제 중 하나는 생성자가 노이즈 같은 이미지를 생성하는 데서 멈추는 것입니다. 판별자와 생성자 양쪽에 모두 드롭아웃을 사용하는 것이 해결 방법이 될 수 있습니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32768)             1081344   \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_1 (LeakyReLU)    (None, 32768)             0         \n",
      "_________________________________________________________________\n",
      "reshape_1 (Reshape)          (None, 16, 16, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 16, 16, 256)       819456    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_2 (LeakyReLU)    (None, 16, 16, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_transpose_1 (Conv2DTr (None, 32, 32, 256)       1048832   \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_3 (LeakyReLU)    (None, 32, 32, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 32, 32, 256)       1638656   \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_4 (LeakyReLU)    (None, 32, 32, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 32, 32, 256)       1638656   \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_5 (LeakyReLU)    (None, 32, 32, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 32, 32, 3)         37635     \n",
      "=================================================================\n",
      "Total params: 6,264,579\n",
      "Trainable params: 6,264,579\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "from keras import layers\n",
    "import numpy as np\n",
    "\n",
    "latent_dim = 32\n",
    "height = 32\n",
    "width = 32\n",
    "channels = 3\n",
    "\n",
    "generator_input = keras.Input(shape=(latent_dim,))\n",
    "\n",
    "# 입력을 16 × 16 크기의 128개 채널을 가진 특성 맵으로 변환합니다\n",
    "x = layers.Dense(128 * 16 * 16)(generator_input)\n",
    "x = layers.LeakyReLU()(x)\n",
    "x = layers.Reshape((16, 16, 128))(x)\n",
    "\n",
    "# 합성곱 층을 추가합니다\n",
    "x = layers.Conv2D(256, 5, padding='same')(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "\n",
    "# 32 × 32 크기로 업샘플링합니다\n",
    "x = layers.Conv2DTranspose(256, 4, strides=2, padding='same')(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "\n",
    "# 합성곱 층을 더 추가합니다\n",
    "x = layers.Conv2D(256, 5, padding='same')(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "x = layers.Conv2D(256, 5, padding='same')(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "\n",
    "# 32 × 32 크기의 1개 채널을 가진 특성 맵을 생성합니다\n",
    "x = layers.Conv2D(channels, 7, activation='tanh', padding='same')(x)\n",
    "generator = keras.models.Model(generator_input, x)\n",
    "generator.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 판별자\n",
    "\n",
    "다음은 후보 이미지(진짜 혹은 가짜)를 입력으로 받고 두 개의 클래스로 분류하는 `discriminator` 모델을 만들겠습니다. 이 클래스는 '생성된 이미지' 또는 '훈련 세트에서 온 진짜 이미지'입니다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 32, 32, 3)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 30, 30, 128)       3584      \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_6 (LeakyReLU)    (None, 30, 30, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 14, 14, 128)       262272    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_7 (LeakyReLU)    (None, 14, 14, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 6, 6, 128)         262272    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_8 (LeakyReLU)    (None, 6, 6, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 2, 2, 128)         262272    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_9 (LeakyReLU)    (None, 2, 2, 128)         0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1)                 513       \n",
      "=================================================================\n",
      "Total params: 790,913\n",
      "Trainable params: 790,913\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "discriminator_input = layers.Input(shape=(height, width, channels))\n",
    "x = layers.Conv2D(128, 3)(discriminator_input)\n",
    "x = layers.LeakyReLU()(x)\n",
    "x = layers.Conv2D(128, 4, strides=2)(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "x = layers.Conv2D(128, 4, strides=2)(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "x = layers.Conv2D(128, 4, strides=2)(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "x = layers.Flatten()(x)\n",
    "\n",
    "# 드롭아웃 층을 넣는 것이 아주 중요합니다!\n",
    "x = layers.Dropout(0.4)(x)\n",
    "\n",
    "# 분류 층\n",
    "x = layers.Dense(1, activation='sigmoid')(x)\n",
    "\n",
    "discriminator = keras.models.Model(discriminator_input, x)\n",
    "discriminator.summary()\n",
    "\n",
    "# 옵티마이저에서 (값을 지정하여) 그래디언트 클리핑을 사용합니다\n",
    "# 안정된 훈련을 위해서 학습률 감쇠를 사용합니다\n",
    "discriminator_optimizer = keras.optimizers.RMSprop(lr=0.0008, clipvalue=1.0, decay=1e-8)\n",
    "discriminator.compile(optimizer=discriminator_optimizer, loss='binary_crossentropy')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 적대적 네트워크\n",
    "\n",
    "마지막으로 생성자와 판별자를 연결하여 GAN을 설정합니다. 훈련할 때 생성자가 판별자를 속이는 능력이 커지도록 학습합니다. 이 모델은 잠재 공간의 포인트를 “진짜\" 또는 “가짜\"의 분류 결정으로 변환합니다. 훈련에 사용되는 타깃 레이블은 항상 '진짜 이미지'입니다. `gan`을 훈련하는 것은 `discriminator`가 가짜 이미지를 보았을 때 진짜라고 예측하도록 만들기 위해 `generator`의 가중치를 업데이트하는 것입니다. 훈련하는 동안 판별자를 동결(학습되지 않도록)하는 것이 아주 중요합니다. `gan`을 훈련할 때 가중치가 업데이트되지 않습니다. 판별자의 가중치가 훈련하는 동안 업데이트되면 판별자는 항상 “진짜\"를 예측하도록 훈련됩니다. 이것이 우리가 원하는 바는 아니죠!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 판별자의 가중치가 훈련되지 않도록 설정합니다(gan 모델에만 적용됩니다)\n",
    "discriminator.trainable = False\n",
    "\n",
    "gan_input = keras.Input(shape=(latent_dim,))\n",
    "gan_output = discriminator(generator(gan_input))\n",
    "gan = keras.models.Model(gan_input, gan_output)\n",
    "\n",
    "gan_optimizer = keras.optimizers.RMSprop(lr=0.0004, clipvalue=1.0, decay=1e-8)\n",
    "gan.compile(optimizer=gan_optimizer, loss='binary_crossentropy')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## DCGAN 훈련 방법\n",
    "\n",
    "이제 훈련을 시작합니다.  훈련 반복의 내용을 요약 정리해 보겠습니다.\n",
    "\n",
    "```\n",
    "매 반복마다 다음을 수행합니다:\n",
    "\n",
    "    1.잠재 공간에서 무작위로 포인트를 뽑습니다(랜덤 노이즈).\n",
    "    2.이 랜덤 노이즈를 사용해 `generator`에서 이미지를 생성합니다.\n",
    "    3.생성된 이미지와 진짜 이미지를 섞습니다.\n",
    "    4.진짜와 가짜가 섞인 이미지와 이에 대응하는 타깃을 사용해 `discriminator`를 훈련합니다. 타깃은 “진짜\"(실제 이미지일 경우) 또는 “가짜\"(생성된 이미지일 경우)입니다.\n",
    "    5.잠재 공간에서 무작위로 새로운 포인트를 뽑습니다.\n",
    "    6.이 랜덤 벡터를 사용해 `gan`을 훈련합니다. 모든 타깃은 “진짜\"로 설정합니다. 판별자가 생성된 이미지를 모두 “진짜 이미지\"라고 예측하도록 생성자의 가중치를 업데이트합니다(`gan` 안에서 판별자는 동결되기 때문에 생성자만 업데이트합니다). 결국 생성자는 판별자를 속이도록 훈련합니다.\n",
    "```\n",
    "\n",
    "실제로 만들어 보죠:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/haesun/anaconda3/envs/deep-learning-with-python/lib/python3.6/site-packages/keras/engine/training.py:490: UserWarning: Discrepancy between trainable weights and collected trainable weights, did you set `model.trainable` without calling `model.compile` after ?\n",
      "  'Discrepancy between trainable weights and collected trainable'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "스텝 0에서 판별자 손실: 0.6859887\n",
      "스텝 0에서 적대적 손실: 0.6648078\n",
      "스텝 100에서 판별자 손실: 0.98458844\n",
      "스텝 100에서 적대적 손실: 15.942385\n",
      "스텝 200에서 판별자 손실: 0.45530224\n",
      "스텝 200에서 적대적 손실: 3.9116611\n",
      "스텝 300에서 판별자 손실: 0.70805806\n",
      "스텝 300에서 적대적 손실: 0.758634\n",
      "스텝 400에서 판별자 손실: 0.7023996\n",
      "스텝 400에서 적대적 손실: 0.76458657\n",
      "스텝 500에서 판별자 손실: 0.66910684\n",
      "스텝 500에서 적대적 손실: 0.8703934\n",
      "스텝 600에서 판별자 손실: 0.7044946\n",
      "스텝 600에서 적대적 손실: 0.7595641\n",
      "스텝 700에서 판별자 손실: 0.70370543\n",
      "스텝 700에서 적대적 손실: 0.7494671\n",
      "스텝 800에서 판별자 손실: 0.6888744\n",
      "스텝 800에서 적대적 손실: 0.72338116\n",
      "스텝 900에서 판별자 손실: 0.70839185\n",
      "스텝 900에서 적대적 손실: 0.95430785\n",
      "스텝 1000에서 판별자 손실: 0.68738854\n",
      "스텝 1000에서 적대적 손실: 0.73358965\n",
      "스텝 1100에서 판별자 손실: 0.6951221\n",
      "스텝 1100에서 적대적 손실: 0.76310116\n",
      "스텝 1200에서 판별자 손실: 0.69487345\n",
      "스텝 1200에서 적대적 손실: 0.75073826\n",
      "스텝 1300에서 판별자 손실: 0.70006037\n",
      "스텝 1300에서 적대적 손실: 0.7711214\n",
      "스텝 1400에서 판별자 손실: 0.69141567\n",
      "스텝 1400에서 적대적 손실: 0.762586\n",
      "스텝 1500에서 판별자 손실: 0.692662\n",
      "스텝 1500에서 적대적 손실: 0.766914\n",
      "스텝 1600에서 판별자 손실: 0.70417833\n",
      "스텝 1600에서 적대적 손실: 0.7402885\n",
      "스텝 1700에서 판별자 손실: 0.7007116\n",
      "스텝 1700에서 적대적 손실: 0.788754\n",
      "스텝 1800에서 판별자 손실: 0.6862414\n",
      "스텝 1800에서 적대적 손실: 0.90469587\n",
      "스텝 1900에서 판별자 손실: 0.68944335\n",
      "스텝 1900에서 적대적 손실: 0.753205\n",
      "스텝 2000에서 판별자 손실: 0.73440367\n",
      "스텝 2000에서 적대적 손실: 0.7881855\n",
      "스텝 2100에서 판별자 손실: 0.68764937\n",
      "스텝 2100에서 적대적 손실: 0.77248967\n",
      "스텝 2200에서 판별자 손실: 0.69836366\n",
      "스텝 2200에서 적대적 손실: 0.77662754\n",
      "스텝 2300에서 판별자 손실: 0.685547\n",
      "스텝 2300에서 적대적 손실: 0.75267434\n",
      "스텝 2400에서 판별자 손실: 0.7113415\n",
      "스텝 2400에서 적대적 손실: 0.7478822\n",
      "스텝 2500에서 판별자 손실: 0.698676\n",
      "스텝 2500에서 적대적 손실: 0.74612296\n",
      "스텝 2600에서 판별자 손실: 0.6922611\n",
      "스텝 2600에서 적대적 손실: 0.7469531\n",
      "스텝 2700에서 판별자 손실: 0.6905508\n",
      "스텝 2700에서 적대적 손실: 0.7141514\n",
      "스텝 2800에서 판별자 손실: 0.6967103\n",
      "스텝 2800에서 적대적 손실: 0.7513309\n",
      "스텝 2900에서 판별자 손실: 0.69222534\n",
      "스텝 2900에서 적대적 손실: 0.784592\n",
      "스텝 3000에서 판별자 손실: 0.6932871\n",
      "스텝 3000에서 적대적 손실: 0.76088995\n",
      "스텝 3100에서 판별자 손실: 0.68727434\n",
      "스텝 3100에서 적대적 손실: 0.6500176\n",
      "스텝 3200에서 판별자 손실: 0.72109354\n",
      "스텝 3200에서 적대적 손실: 0.7601031\n",
      "스텝 3300에서 판별자 손실: 0.69486094\n",
      "스텝 3300에서 적대적 손실: 0.7491412\n",
      "스텝 3400에서 판별자 손실: 0.6895932\n",
      "스텝 3400에서 적대적 손실: 0.7161566\n",
      "스텝 3500에서 판별자 손실: 0.6968385\n",
      "스텝 3500에서 적대적 손실: 0.7591454\n",
      "스텝 3600에서 판별자 손실: 0.6983391\n",
      "스텝 3600에서 적대적 손실: 0.75471365\n",
      "스텝 3700에서 판별자 손실: 0.6767294\n",
      "스텝 3700에서 적대적 손실: 0.8817333\n",
      "스텝 3800에서 판별자 손실: 0.6919452\n",
      "스텝 3800에서 적대적 손실: 0.75868195\n",
      "스텝 3900에서 판별자 손실: 0.7031524\n",
      "스텝 3900에서 적대적 손실: 0.7505781\n",
      "스텝 4000에서 판별자 손실: 0.6838635\n",
      "스텝 4000에서 적대적 손실: 0.75076956\n",
      "스텝 4100에서 판별자 손실: 0.6887736\n",
      "스텝 4100에서 적대적 손실: 0.7528058\n",
      "스텝 4200에서 판별자 손실: 0.69331276\n",
      "스텝 4200에서 적대적 손실: 0.79471785\n",
      "스텝 4300에서 판별자 손실: 0.68512636\n",
      "스텝 4300에서 적대적 손실: 0.75867164\n",
      "스텝 4400에서 판별자 손실: 0.6883174\n",
      "스텝 4400에서 적대적 손실: 0.7720972\n",
      "스텝 4500에서 판별자 손실: 0.7012172\n",
      "스텝 4500에서 적대적 손실: 0.67885685\n",
      "스텝 4600에서 판별자 손실: 0.69478226\n",
      "스텝 4600에서 적대적 손실: 0.7391167\n",
      "스텝 4700에서 판별자 손실: 0.70749074\n",
      "스텝 4700에서 적대적 손실: 0.8058073\n",
      "스텝 4800에서 판별자 손실: 0.7214985\n",
      "스텝 4800에서 적대적 손실: 1.0094569\n",
      "스텝 4900에서 판별자 손실: 0.6968532\n",
      "스텝 4900에서 적대적 손실: 0.76501334\n",
      "스텝 5000에서 판별자 손실: 0.70068586\n",
      "스텝 5000에서 적대적 손실: 0.8418042\n",
      "스텝 5100에서 판별자 손실: 0.6901635\n",
      "스텝 5100에서 적대적 손실: 0.73663354\n",
      "스텝 5200에서 판별자 손실: 0.69752437\n",
      "스텝 5200에서 적대적 손실: 0.72524816\n",
      "스텝 5300에서 판별자 손실: 0.6945907\n",
      "스텝 5300에서 적대적 손실: 0.7393595\n",
      "스텝 5400에서 판별자 손실: 0.69078934\n",
      "스텝 5400에서 적대적 손실: 0.74214107\n",
      "스텝 5500에서 판별자 손실: 0.6939332\n",
      "스텝 5500에서 적대적 손실: 0.7277419\n",
      "스텝 5600에서 판별자 손실: 0.6989276\n",
      "스텝 5600에서 적대적 손실: 0.74151313\n",
      "스텝 5700에서 판별자 손실: 0.6865061\n",
      "스텝 5700에서 적대적 손실: 0.7271993\n",
      "스텝 5800에서 판별자 손실: 0.7233262\n",
      "스텝 5800에서 적대적 손실: 0.99431247\n",
      "스텝 5900에서 판별자 손실: 0.70499647\n",
      "스텝 5900에서 적대적 손실: 1.0466232\n",
      "스텝 6000에서 판별자 손실: 0.6851308\n",
      "스텝 6000에서 적대적 손실: 0.7371407\n",
      "스텝 6100에서 판별자 손실: 0.68779993\n",
      "스텝 6100에서 적대적 손실: 0.7095089\n",
      "스텝 6200에서 판별자 손실: 0.69324887\n",
      "스텝 6200에서 적대적 손실: 0.7428337\n",
      "스텝 6300에서 판별자 손실: 0.68435943\n",
      "스텝 6300에서 적대적 손실: 0.6888761\n",
      "스텝 6400에서 판별자 손실: 0.7009553\n",
      "스텝 6400에서 적대적 손실: 0.7570043\n",
      "스텝 6500에서 판별자 손실: 0.69860137\n",
      "스텝 6500에서 적대적 손실: 0.8652438\n",
      "스텝 6600에서 판별자 손실: 0.6948483\n",
      "스텝 6600에서 적대적 손실: 0.81759083\n",
      "스텝 6700에서 판별자 손실: 0.6920208\n",
      "스텝 6700에서 적대적 손실: 0.71276224\n",
      "스텝 6800에서 판별자 손실: 0.7139355\n",
      "스텝 6800에서 적대적 손실: 0.7371713\n",
      "스텝 6900에서 판별자 손실: 0.6853081\n",
      "스텝 6900에서 적대적 손실: 0.73708713\n",
      "스텝 7000에서 판별자 손실: 0.69366324\n",
      "스텝 7000에서 적대적 손실: 0.79389805\n",
      "스텝 7100에서 판별자 손실: 0.7064244\n",
      "스텝 7100에서 적대적 손실: 0.7813763\n",
      "스텝 7200에서 판별자 손실: 0.6926437\n",
      "스텝 7200에서 적대적 손실: 0.7462541\n",
      "스텝 7300에서 판별자 손실: 0.699135\n",
      "스텝 7300에서 적대적 손실: 0.75480306\n",
      "스텝 7400에서 판별자 손실: 0.69808084\n",
      "스텝 7400에서 적대적 손실: 0.7792807\n",
      "스텝 7500에서 판별자 손실: 0.6886368\n",
      "스텝 7500에서 적대적 손실: 0.7415221\n",
      "스텝 7600에서 판별자 손실: 0.698572\n",
      "스텝 7600에서 적대적 손실: 0.77041376\n",
      "스텝 7700에서 판별자 손실: 0.6963496\n",
      "스텝 7700에서 적대적 손실: 0.7734369\n",
      "스텝 7800에서 판별자 손실: 0.6873445\n",
      "스텝 7800에서 적대적 손실: 0.78740144\n",
      "스텝 7900에서 판별자 손실: 0.6971999\n",
      "스텝 7900에서 적대적 손실: 0.78107464\n",
      "스텝 8000에서 판별자 손실: 0.6985973\n",
      "스텝 8000에서 적대적 손실: 0.7444295\n",
      "스텝 8100에서 판별자 손실: 0.6955716\n",
      "스텝 8100에서 적대적 손실: 0.7179367\n",
      "스텝 8200에서 판별자 손실: 0.69686717\n",
      "스텝 8200에서 적대적 손실: 0.86814547\n",
      "스텝 8300에서 판별자 손실: 0.7090989\n",
      "스텝 8300에서 적대적 손실: 0.8461113\n",
      "스텝 8400에서 판별자 손실: 0.6789621\n",
      "스텝 8400에서 적대적 손실: 0.7891714\n",
      "스텝 8500에서 판별자 손실: 0.6828822\n",
      "스텝 8500에서 적대적 손실: 0.744535\n",
      "스텝 8600에서 판별자 손실: 0.73256254\n",
      "스텝 8600에서 적대적 손실: 0.8267523\n",
      "스텝 8700에서 판별자 손실: 0.6953441\n",
      "스텝 8700에서 적대적 손실: 0.83365214\n",
      "스텝 8800에서 판별자 손실: 0.67792505\n",
      "스텝 8800에서 적대적 손실: 0.7299119\n",
      "스텝 8900에서 판별자 손실: 0.69003236\n",
      "스텝 8900에서 적대적 손실: 0.77839434\n",
      "스텝 9000에서 판별자 손실: 0.68872595\n",
      "스텝 9000에서 적대적 손실: 0.7916759\n",
      "스텝 9100에서 판별자 손실: 0.6975969\n",
      "스텝 9100에서 적대적 손실: 0.7680103\n",
      "스텝 9200에서 판별자 손실: 0.68702763\n",
      "스텝 9200에서 적대적 손실: 0.7282267\n",
      "스텝 9300에서 판별자 손실: 0.7355779\n",
      "스텝 9300에서 적대적 손실: 0.8837675\n",
      "스텝 9400에서 판별자 손실: 0.68756074\n",
      "스텝 9400에서 적대적 손실: 0.9133609\n",
      "스텝 9500에서 판별자 손실: 0.6581781\n",
      "스텝 9500에서 적대적 손실: 0.75360084\n",
      "스텝 9600에서 판별자 손실: 0.6929166\n",
      "스텝 9600에서 적대적 손실: 0.8556153\n",
      "스텝 9700에서 판별자 손실: 0.6904553\n",
      "스텝 9700에서 적대적 손실: 1.0184231\n",
      "스텝 9800에서 판별자 손실: 0.65708625\n",
      "스텝 9800에서 적대적 손실: 0.6254333\n",
      "스텝 9900에서 판별자 손실: 0.68810195\n",
      "스텝 9900에서 적대적 손실: 0.884272\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "from keras.preprocessing import image\n",
    "\n",
    "# CIFAR10 데이터를 로드합니다\n",
    "(x_train, y_train), (_, _) = keras.datasets.cifar10.load_data()\n",
    "\n",
    "# 개구리 이미지를 선택합니다(클래스 6)\n",
    "x_train = x_train[y_train.flatten() == 6]\n",
    "\n",
    "# 데이터를 정규화합니다\n",
    "x_train = x_train.reshape(\n",
    "    (x_train.shape[0],) + (height, width, channels)).astype('float32') / 255.\n",
    "\n",
    "iterations = 10000\n",
    "batch_size = 20\n",
    "save_dir = './datasets/gan_images/'\n",
    "if not os.path.exists(save_dir):\n",
    "    os.mkdir(save_dir)\n",
    "\n",
    "# 훈련 반복 시작\n",
    "start = 0\n",
    "for step in range(iterations):\n",
    "    # 잠재 공간에서 무작위로 포인트를 샘플링합니다\n",
    "    random_latent_vectors = np.random.normal(size=(batch_size, latent_dim))\n",
    "\n",
    "    # 가짜 이미지를 디코딩합니다\n",
    "    generated_images = generator.predict(random_latent_vectors)\n",
    "\n",
    "    # 진짜 이미지와 연결합니다\n",
    "    stop = start + batch_size\n",
    "    real_images = x_train[start: stop]\n",
    "    combined_images = np.concatenate([generated_images, real_images])\n",
    "\n",
    "    # 진짜와 가짜 이미지를 구분하여 레이블을 합칩니다\n",
    "    labels = np.concatenate([np.ones((batch_size, 1)),\n",
    "                             np.zeros((batch_size, 1))])\n",
    "    # 레이블에 랜덤 노이즈를 추가합니다. 아주 중요합니다!\n",
    "    labels += 0.05 * np.random.random(labels.shape)\n",
    "\n",
    "    # discriminator를 훈련합니다\n",
    "    d_loss = discriminator.train_on_batch(combined_images, labels)\n",
    "\n",
    "    # 잠재 공간에서 무작위로 포인트를 샘플링합니다\n",
    "    random_latent_vectors = np.random.normal(size=(batch_size, latent_dim))\n",
    "\n",
    "    # 모두 “진짜 이미지\"라고 레이블을 만듭니다\n",
    "    misleading_targets = np.zeros((batch_size, 1))\n",
    "\n",
    "    # generator를 훈련합니다(gan 모델에서 discriminator의 가중치는 동결됩니다)\n",
    "    a_loss = gan.train_on_batch(random_latent_vectors, misleading_targets)\n",
    "    \n",
    "    start += batch_size\n",
    "    if start > len(x_train) - batch_size:\n",
    "      start = 0\n",
    "\n",
    "    # 중간 중간 저장하고 그래프를 그립니다\n",
    "    if step % 100 == 0:\n",
    "        # 모델 가중치를 저장합니다\n",
    "        gan.save_weights('gan.h5')\n",
    "\n",
    "        # 측정 지표를 출력합니다\n",
    "        print('스텝 %s에서 판별자 손실: %s' % (step, d_loss))\n",
    "        print('스텝 %s에서 적대적 손실: %s' % (step, a_loss))\n",
    "\n",
    "        # 생성된 이미지 하나를 저장합니다\n",
    "        img = image.array_to_img(generated_images[0] * 255., scale=False)\n",
    "        img.save(os.path.join(save_dir, 'generated_frog' + str(step) + '.png'))\n",
    "\n",
    "        # 비교를 위해 진짜 이미지 하나를 저장합니다\n",
    "        img = image.array_to_img(real_images[0] * 255., scale=False)\n",
    "        img.save(os.path.join(save_dir, 'real_frog' + str(step) + '.png'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "가짜 이미지 몇개를 출력해 보죠:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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"
    },
    {
     "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"
    },
    {
     "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"
    },
    {
     "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"
    },
    {
     "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": [
    "# 잠재 공간에서 랜덤한 포인트를 샘플링합니다\n",
    "random_latent_vectors = np.random.normal(size=(10, latent_dim))\n",
    "\n",
    "# 가짜 이미지로 디코딩합니다\n",
    "generated_images = generator.predict(random_latent_vectors)\n",
    "\n",
    "for i in range(generated_images.shape[0]):\n",
    "    img = image.array_to_img(generated_images[i] * 255., scale=False)\n",
    "    plt.figure()\n",
    "    plt.imshow(img)\n",
    "    \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "픽셀 경계가 두드러진 개구리처럼 보이는 이미지를 얻었습니다."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}