16_Generation_GAN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "GAN 은 생성된 이미지가 실제 이미지와 통계적으로 거의 구분이 되지 않도록 강제하여 아주 실제 같은 합성 이미지를 생성한다. \n",
    "\n",
    "GAN을 직관적으로 이해하는 방법은 가짜 피카소 그림을 만드는 위조범을 생각하는 것이다. 처음에 위조범은 형편없이 위조한다. 진짜 피카소 그림과 위조품을 섞어서 그림 판매상에게 보여준다. 판매상은 각 그림이 진짜인지 평가하고 어떤 것이 피카소 그림 같은지 위조범에게 피드백을 준다. 위조범은 자신의 화실로 돌아가 새로운 위조품을 준비한다. 시간이 지남에 따라 위조범은 피카소의 스타일을 모방하는 데 점점 능숙해진다. 그림 판매상은 위조품을 구분하는 데 점점 더 전문가가 된다. 결국 아주 휼륭한 피카소 위조품을 만들어 낼 것이다.\n",
    "\n",
    "위조범 네트워크와 전문가 네트워크가 바로 GAN이다. 두 네트워크는 상대를 이기기 위해 훈련한다. GAN의 네트워크는 2개는 다음과 같다.\n",
    "\n",
    "- 생성자 네트워크(Generator network) : 랜덤 벡터(잠재 공간의 무작위한 포인트)를 입력으로 받아 이를 합성된 이미지로 디코딩한다.\n",
    "- 판별자 네트워크(Discriminartor network)(또는 상대 네트워크) : 이미지(실제 또는 합성 이미지)를 입력으로 받아 훈련 세트에서 온 이미지인지, 생성자 네트워크가 만든 이미지인지 판별한다.\n",
    "\n",
    "생성자 네트워크는 판별자 네트워크를 속이도록 훈련한다. 훈련이 계속될수록 점점 더 실제와 같은 이미지를 생성하게 된다. 실제 이미지와 구분할 수 없는 인공적인 이미지를 만들어 판별자 네트워크가 두 이미지를 동일하게 보도록 만든다. 한편 판별자 네트워크는 생성된 이미지가 실제인지 판별하는 기준을 설정하면서 생성자의 능력 향상에 적응해 간다. 훈련이 끝나면 생성자는 입력 공간에 있는 어떤 포인트를 그럴듯한 이밎로 변환한다. VAE와 달리 이 잠재 공간은 의미 있는 구조를 보장하지 않는다. 특히 이 공간은 연속적이지 않다."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "놀랍게도 GAN은 최적화의 최솟값이 고정되지 않은 시스템이다. 이 책에서 다루는 어떤 훈련 설정과도 다르다. 보통 경사 하강법은 고정된 손실 공간에서 언덕을 내려온느 방법이다. GAN에서는 언덕을 내려오는 매 단계가 조금씩 전체 공간을 바꾼다. 최적화 과정이 최솟값을 찾는 것이 아니라 두 힘 간의 평형점을 찾는 다이나믹 시스템이다. 이런 이유로 GAN은 훈련하기 어렵기로 유명하다. GAN을 만들려면 모델 구조와 훈련 파라미터를 주의 깊게 많이 조정해야한다."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![](http://www.miketyka.com/projects/faces/00128000.jpg.cover.jpg)\n",
    "\n",
    "![](http://www.miketyka.com/projects/faces/00734000.jpg.cover.jpg)\n",
    "\n",
    "> http://www.miketyka.com/?s=faces 에서 보는 GAN으로 만든 이미지"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## GAN 구현 방법"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "간단한 형태의 GAN을 설명하겠다. 구체적인 구현은 심층 합성곱 GAN(DCGAN)을 구현할 것이다. 생성자와 판별자가 심층 컨브넷이다. 특히 생성자에서 이미지 업샘플링을 위해서 Conv2DTranspose 층을 사용한다.\n",
    "\n",
    "CIFAR10 데이터셋의 이미지로 GAN을 훈련하겠다. 이 데이터셋은 32X32 크기의 RGB 이미지 5만 개로 이루어져 있고 10개의 클래스를 가진다. (클래스마다 5,000개의 이미지가 있다.) 문제를 간단하게 만들기 위해 \"frog\"클래스의 이미지만 사용하겠다.\n",
    "\n",
    "GAN의 구조는 다음과 같다.\n",
    "\n",
    "1. generator 네트워크는 (latent_dim), 크기의 벡터를 (32, 32, 3) 크기의 이미지로 매핑한다.\n",
    "2. discriminator 네트워크는 (32, 32, 3) 크기의 이미지가 진짜일 확률을 추정하여 이진 값으로 매핑한다.\n",
    "3. 생성자와 판별자를 연결하는 gan 네트워크를 만든다. gan(x) = discriminator(generator(x))이다. 이 gan 네트워크는 잠재 공간의 벡터를 판별자의 평가로 매핑한다. 즉 판별자는 생성자가 잠재 공간의 벡터를 디코딩한 것이 얼마나 현실적인지 평가한다. \n",
    "4. \"진짜\"/\"가짜\" 레이블과 함께 진짜 이미지와 가자 이미지 샘플을 사용하여 판별자를 훈련한다. 일반적인 이미지 분류 모델을 훈련하는 것과 동일하다.\n",
    "5. 생성자를 훈련하려면 gan 모델의 손실에 대해 생성자 가중치의 그래디언트를 사용한다. 이 말은 매 단계마다 생성자에 의해 디코딩된 이미지를 판별자가 \"진짜\"로 분류하도록 만드는 방향으로 생성자의 가중치를 이동한다는 뜻이다. 다른 말로 하면 판별자를 속이도록 생성자를 훈련한다."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 훈련 방법"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "GAN을 훈련하고 튜닝하는 과정은 어렵기로 유명하다. 알아 두어야 할 몇 가지 유용한 기법이 있다. 딥러닝의 대부분이 그렇듯이 이는 과학보다는 연금술에 가깝다. 이런 기법들은 이론에 바탕을 둔 지침이 아니고 경험을 통해 발견된 것이다. 실제 일어난 현상을 직관적으로 이해하는 수준에서 검증되었다. 모든 문제에 반드시 적용해야 하는 것은 아니지만 경험상 잘 작동한다고 알려져 있다.\n",
    "\n",
    "다음은 이 절에서 GAN 생성자와 판별자를 구현하는 데 사용할 몇 가지 기법이다. 이 목록이 GAN에 관련된 전체 팁이 아니다. GAN 논문들에서 더 많은 방법을 볼 수 있다.\n",
    "\n",
    "- 생성자의 마지막 활성화로 다른 종류의 모델에서 널리 사용되는 sigmoid 대신 tanh 함수를 사용한다.\n",
    "- 균등 분포가 아니고 정규 분포(가우시안 분포)를 사용하여 잠재 공간에서 포인트를 샘플링한다.\n",
    "- 무작위성은 모델을 견고하게 만든다. GAN 훈련은 동적 평형을 만들기 때문에 여러 방식으로 갇힐 가능성이 높다. 훈련하는 동안 무작위성을 주입하면 이를 방지하는 데 도움이 된다. 무작위성은 두 가지 방법으로 주입한다. 판별자에 드롭아웃을 사용하거나 판별자를 위해 레이블에 랜덤 노이즈를 추가한다.\n",
    "- 희소한 그래디언트는 GAN 훈련을 방해할 수 있다. 딥러닝에서 희소는 종종 바람직한 현상이지만 GAN에서는 그렇지 않다. 그래디언트를 희소하게 만들 수 있는 것은 최대풀링 연산과 ReLUJ 활성화 두 가지이다. 최대 풀링 대신 스트라이드 합성곱을 사용하여 다운샘플링을 하는 것이 좋다. 또 ReLU 활성화 대신 LeakyReLU 층을 사용하라. ReLU와 비슷하지만 음수의 활성화 값을 조금 허용하기 때문에 희소가 조금 완화된다.\n",
    "- 생성자에서 픽셀 공간을 균일하게 다루지 못하여 생성된 이미지에서 체스판 모양이 종종 나타난다. 이를 해결하기 위해 생성자와 판별자에서 스트라이드 Conv2DTranpose나 Conv2D를 사용할 때 스트라이드 크기로 나누어질 수 있는 커널 크기를 사용한다."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 생성자"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "먼지 벡터(훈련하는 동안 잠재 공간에서 무작위로 샘플링된다)를 후보 이미지로 변환하는 generator 모델을 만들어 보자. GAN에서 발생하는 많은 문제 중 하나는 생성자 양쪽에 모두 드롭아웃을 사용하는 것이 해결 방법이 될 수 있다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras import layersrs\n",
    "import numpy as np"
   ]
  },
  {
   "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": [
    "latent_dim = 32\n",
    "height = 32\n",
    "width  = 32\n",
    "channels = 3\n",
    "\n",
    "generator_input = keras.Input(shape=(latent_dim,))\n",
    "\n",
    "x = layers.Dense(128 * 16 * 16)(generator_input)\n",
    "x = layers.LeakyReLU()(x)\n",
    "x = layers.Reshape((16, 16, 128))(x)\n",
    "\n",
    "x = layers.Conv2D(256, 5, padding='same')(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "\n",
    "x = layers.Conv2DTranspose(256, 4, strides=2, padding='same')(x)\n",
    "x = layers.LeakyReLU()(x)\n",
    "\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",
    "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"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "다음은 후보 이미지(진짜 또는 가짜)를 입력으로 받고 2개의 클래스로 분류하는 discriminator 모델을 만들겠다. 이 클래스는 '생성된 이미지'또는 '훈련 세트에서 온 진짜 이미지'이다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_3 (InputLayer)         (None, 32, 32, 3)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_9 (Conv2D)            (None, 30, 30, 128)       3584      \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_10 (LeakyReLU)   (None, 30, 30, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_10 (Conv2D)           (None, 14, 14, 128)       262272    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_11 (LeakyReLU)   (None, 14, 14, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_11 (Conv2D)           (None, 6, 6, 128)         262272    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_12 (LeakyReLU)   (None, 6, 6, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_12 (Conv2D)           (None, 2, 2, 128)         262272    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_13 (LeakyReLU)   (None, 2, 2, 128)         0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (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",
    "x = layers.Dropout(0.4)(x)\n",
    "\n",
    "x = layers.Dense(1, activation='sigmoid')(x)\n",
    "\n",
    "discriminator = keras.models.Model(discriminator_input, x)\n",
    "discriminator.summary()\n",
    "\n",
    "discriminator_optimizer = keras.optimizers.RMSprop(\n",
    "    lr=0.0008,\n",
    "    clipvalue=1.0,\n",
    "    decay=1e-8)\n",
    "discriminator.compile(optimizer=discriminator_optimizer,\n",
    "                      loss='binary_crossentropy')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 적대적 네트워크"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "마지막으로 생성자와 판별자를 연결하여 GAN을 설정한다. 훈련할 때 생성자가 판별자를 속이는 능력이 커지도록 학습한다. 이 모델은 잠재 공간의 포인트를 \"진짜\" 또는 \"가짜\"의 분류 결정으로 변환한다. 훈련에 사용되는 타깃 레이블은 항상 '진짜 이미지'이다. GAN을 훈련하는 것은 discriminator가 가짜 이미지를 보았을 때 진짜라고 예측하도록 만들기 위해 generator의 가중치를 업데이트 하는 것이다. 훈련하는 동안 판별자를 동결(학습되지 않도록)하는 것이 아주 중요하다. GAN을 훈련할 때 가중치가 업데이트되지 않는다. 판별자의 가중치가 훈련하는 동안 업데이트되면 판별자는 항상 \"진짜\"를 예측하도록 훈련된다. 이것은 우리가 원하는 바가 아니다!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "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"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "이제 훈련을 시작한다. 훈련 반복의 내용을 요약 정리해 보겠다. 매 반복마다 다음을 수행한다.\n",
    "\n",
    "1. 잠재 공간에서 무작위로 포인트를 뽑는다(랜덤 노이즈)\n",
    "2. 이 랜덤 노이즈를 사용하여 generator에서 이미지를 생성한다.\n",
    "3. 생성된 이미지와 진짜 이미지를 섞는다.\n",
    "4. 진짜와 가짜가 섞인 이미지와 이에 대응하는 타깃을 사용하여 discriminator를 훈련한다. 타깃은 \"진짜(실제 이미지일 경우)\" 또는 \"가짜(생성된 이미지일 경우)\"이다.\n",
    "5. 잠재 공간에서 무작위로 새로운 포인트를 뽑는다.\n",
    "6. 이 랜덤 벡터를 사용하여 gan을 훈련한다. 모든 타깃은 \"진짜\"로 설정한다. 판별자가 생성된 이미지를 모두 \"진짜 이미지\"라고 예측하도록 생성자의 가중치를 업데이트한다(gan)안에서 판별자는 동결되기 때문에 생성자만 업데이트한다) 결국 생성자는 판별자를 속이도록 훈련한다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "from keras.preprocessing import image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Anaconda3\\envs\\tf_gpu\\lib\\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": [
      "discriminator loss :  0.6883102\n",
      "generative loss :  0.6883102\n",
      "discriminator loss :  0.7079845\n",
      "generative loss :  0.7079845\n",
      "discriminator loss :  0.6811552\n",
      "generative loss :  0.6811552\n",
      "discriminator loss :  0.68777305\n",
      "generative loss :  0.68777305\n",
      "discriminator loss :  0.68104184\n",
      "generative loss :  0.68104184\n",
      "discriminator loss :  0.6952647\n",
      "generative loss :  0.6952647\n",
      "discriminator loss :  0.6896124\n",
      "generative loss :  0.6896124\n",
      "discriminator loss :  0.69773656\n",
      "generative loss :  0.69773656\n",
      "discriminator loss :  0.691177\n",
      "generative loss :  0.691177\n",
      "discriminator loss :  0.69689035\n",
      "generative loss :  0.69689035\n",
      "discriminator loss :  0.6939559\n",
      "generative loss :  0.6939559\n",
      "discriminator loss :  0.6933508\n",
      "generative loss :  0.6933508\n",
      "discriminator loss :  0.7039848\n",
      "generative loss :  0.7039848\n",
      "discriminator loss :  0.69345075\n",
      "generative loss :  0.69345075\n",
      "discriminator loss :  0.69653356\n",
      "generative loss :  0.69653356\n",
      "discriminator loss :  0.7783613\n",
      "generative loss :  0.7783613\n",
      "discriminator loss :  0.683427\n",
      "generative loss :  0.683427\n",
      "discriminator loss :  0.70361245\n",
      "generative loss :  0.70361245\n",
      "discriminator loss :  0.7001464\n",
      "generative loss :  0.7001464\n",
      "discriminator loss :  0.6914102\n",
      "generative loss :  0.6914102\n",
      "discriminator loss :  0.69378585\n",
      "generative loss :  0.69378585\n",
      "discriminator loss :  0.6970881\n",
      "generative loss :  0.6970881\n",
      "discriminator loss :  0.7384157\n",
      "generative loss :  0.7384157\n",
      "discriminator loss :  0.688681\n",
      "generative loss :  0.688681\n",
      "discriminator loss :  0.73175704\n",
      "generative loss :  0.73175704\n",
      "discriminator loss :  0.6950456\n",
      "generative loss :  0.6950456\n",
      "discriminator loss :  0.6897127\n",
      "generative loss :  0.6897127\n",
      "discriminator loss :  0.7002765\n",
      "generative loss :  0.7002765\n",
      "discriminator loss :  0.6907035\n",
      "generative loss :  0.6907035\n",
      "discriminator loss :  0.69561225\n",
      "generative loss :  0.69561225\n",
      "discriminator loss :  0.689071\n",
      "generative loss :  0.689071\n",
      "discriminator loss :  0.6771034\n",
      "generative loss :  0.6771034\n",
      "discriminator loss :  0.7033454\n",
      "generative loss :  0.7033454\n",
      "discriminator loss :  0.81828463\n",
      "generative loss :  0.81828463\n",
      "discriminator loss :  0.6909939\n",
      "generative loss :  0.6909939\n",
      "discriminator loss :  0.6831749\n",
      "generative loss :  0.6831749\n",
      "discriminator loss :  0.6846794\n",
      "generative loss :  0.6846794\n",
      "discriminator loss :  0.70465636\n",
      "generative loss :  0.70465636\n",
      "discriminator loss :  0.6721223\n",
      "generative loss :  0.6721223\n",
      "discriminator loss :  0.6944953\n",
      "generative loss :  0.6944953\n",
      "discriminator loss :  0.7430039\n",
      "generative loss :  0.7430039\n",
      "discriminator loss :  0.7001158\n",
      "generative loss :  0.7001158\n",
      "discriminator loss :  0.7008239\n",
      "generative loss :  0.7008239\n",
      "discriminator loss :  0.68723404\n",
      "generative loss :  0.68723404\n",
      "discriminator loss :  0.7017851\n",
      "generative loss :  0.7017851\n",
      "discriminator loss :  0.7080625\n",
      "generative loss :  0.7080625\n",
      "discriminator loss :  0.69898593\n",
      "generative loss :  0.69898593\n",
      "discriminator loss :  0.69379175\n",
      "generative loss :  0.69379175\n",
      "discriminator loss :  0.71090084\n",
      "generative loss :  0.71090084\n",
      "discriminator loss :  0.69443053\n",
      "generative loss :  0.69443053\n",
      "discriminator loss :  0.6781335\n",
      "generative loss :  0.6781335\n",
      "discriminator loss :  0.69231653\n",
      "generative loss :  0.69231653\n",
      "discriminator loss :  0.69943714\n",
      "generative loss :  0.69943714\n",
      "discriminator loss :  0.69474363\n",
      "generative loss :  0.69474363\n",
      "discriminator loss :  0.67507\n",
      "generative loss :  0.67507\n",
      "discriminator loss :  0.6912862\n",
      "generative loss :  0.6912862\n",
      "discriminator loss :  0.7080633\n",
      "generative loss :  0.7080633\n",
      "discriminator loss :  0.6981883\n",
      "generative loss :  0.6981883\n",
      "discriminator loss :  0.69721687\n",
      "generative loss :  0.69721687\n",
      "discriminator loss :  0.74897075\n",
      "generative loss :  0.74897075\n",
      "discriminator loss :  0.68598795\n",
      "generative loss :  0.68598795\n",
      "discriminator loss :  0.7171116\n",
      "generative loss :  0.7171116\n",
      "discriminator loss :  0.74244195\n",
      "generative loss :  0.74244195\n",
      "discriminator loss :  0.6521789\n",
      "generative loss :  0.6521789\n",
      "discriminator loss :  0.710044\n",
      "generative loss :  0.710044\n",
      "discriminator loss :  0.68024063\n",
      "generative loss :  0.68024063\n",
      "discriminator loss :  0.72785616\n",
      "generative loss :  0.72785616\n",
      "discriminator loss :  0.75283813\n",
      "generative loss :  0.75283813\n",
      "discriminator loss :  0.82009715\n",
      "generative loss :  0.82009715\n",
      "discriminator loss :  0.66917515\n",
      "generative loss :  0.66917515\n",
      "discriminator loss :  0.6636988\n",
      "generative loss :  0.6636988\n",
      "discriminator loss :  0.6898284\n",
      "generative loss :  0.6898284\n",
      "discriminator loss :  0.68446434\n",
      "generative loss :  0.68446434\n",
      "discriminator loss :  0.648232\n",
      "generative loss :  0.648232\n",
      "discriminator loss :  0.68009347\n",
      "generative loss :  0.68009347\n",
      "discriminator loss :  0.6676814\n",
      "generative loss :  0.6676814\n",
      "discriminator loss :  0.6695528\n",
      "generative loss :  0.6695528\n",
      "discriminator loss :  0.68657744\n",
      "generative loss :  0.68657744\n",
      "discriminator loss :  0.70213234\n",
      "generative loss :  0.70213234\n",
      "discriminator loss :  0.6617445\n",
      "generative loss :  0.6617445\n",
      "discriminator loss :  0.6508812\n",
      "generative loss :  0.6508812\n",
      "discriminator loss :  0.72824645\n",
      "generative loss :  0.72824645\n",
      "discriminator loss :  0.6875546\n",
      "generative loss :  0.6875546\n",
      "discriminator loss :  0.6780707\n",
      "generative loss :  0.6780707\n",
      "discriminator loss :  0.671913\n",
      "generative loss :  0.671913\n",
      "discriminator loss :  0.7122441\n",
      "generative loss :  0.7122441\n",
      "discriminator loss :  0.6950568\n",
      "generative loss :  0.6950568\n",
      "discriminator loss :  0.69495785\n",
      "generative loss :  0.69495785\n",
      "discriminator loss :  0.78331584\n",
      "generative loss :  0.78331584\n",
      "discriminator loss :  0.66571236\n",
      "generative loss :  0.66571236\n",
      "discriminator loss :  0.7092401\n",
      "generative loss :  0.7092401\n",
      "discriminator loss :  0.6733028\n",
      "generative loss :  0.6733028\n",
      "discriminator loss :  0.74558574\n",
      "generative loss :  0.74558574\n",
      "discriminator loss :  0.5889947\n",
      "generative loss :  0.5889947\n",
      "discriminator loss :  0.68434584\n",
      "generative loss :  0.68434584\n",
      "discriminator loss :  0.6750849\n",
      "generative loss :  0.6750849\n",
      "discriminator loss :  0.7036767\n",
      "generative loss :  0.7036767\n",
      "discriminator loss :  0.7089672\n",
      "generative loss :  0.7089672\n",
      "discriminator loss :  0.6624735\n",
      "generative loss :  0.6624735\n",
      "discriminator loss :  0.64132607\n",
      "generative loss :  0.64132607\n"
     ]
    }
   ],
   "source": [
    "(x_train, y_train), (_, _) = keras.datasets.cifar10.load_data()\n",
    "\n",
    "x_train = x_train[y_train.flatten() == 6]\n",
    "\n",
    "x_train = x_train.reshape(\n",
    "    (x_train.shape[0],) + \n",
    "    (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",
    "start = 0\n",
    "for step in range(iterations):\n",
    "    random_latent_vectors = np.random.normal(size=(batch_size,\n",
    "                                                   latent_dim))\n",
    "    generated_images = generator.predict(random_latent_vectors)\n",
    "    \n",
    "    stop = start + batch_size\n",
    "    real_images = x_train[start : stop]\n",
    "    combined_images = np.concatenate([generated_images, real_images])\n",
    "    \n",
    "    labels = np.concatenate([np.ones((batch_size, 1)),\n",
    "                             np.zeros((batch_size, 1))])\n",
    "    labels += 0.05 * np.random.random(labels.shape)\n",
    "    \n",
    "    d_loss = discriminator.train_on_batch(combined_images, labels)\n",
    "    \n",
    "    random_latent_vectors = np.random.normal(size=(batch_size,\n",
    "                                                   latent_dim))\n",
    "    \n",
    "    misleading_targets = np.zeros((batch_size, 1))\n",
    "    \n",
    "    a_loss = gan.train_on_batch(random_latent_vectors,\n",
    "                                misleading_targets)\n",
    "    \n",
    "    start += batch_size\n",
    "    if start > len(x_train) - batch_size:\n",
    "        start = 0\n",
    "    if step % 100 == 0:\n",
    "        gan.save_weights('gan.h5')\n",
    "        \n",
    "        print('discriminator loss : ', d_loss)\n",
    "        print('generative loss : ', d_loss)\n",
    "        \n",
    "        img = image.array_to_img(generated_images[0] * 255., scale=False)\n",
    "        img.save(os.path.join(save_dir,\n",
    "                             'generated_frog' + str(step) + '.png'))\n",
    "        \n",
    "        img = image.array_to_img(real_images[0] * 255., scale=False)\n",
    "        img.save(os.path.join(save_dir,\n",
    "                             'real_frog'+str(step) + '.png'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "훈련할 때 적대적 손실이 크게 증가하는 것을 볼지도 모른다. 반면에 판별자의 손실은 0으로 향한다. 결국 판별자가 생성자를 압도할 수 있다."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": false
   },
   "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": [
    "random_latent_vectors = np.random.normal(size=(10, latent_dim))\n",
    "\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": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf_gpu",
   "language": "python",
   "name": "tf_gpu"
  },
  "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
}