1、理论讲解,清晰易懂:
一文看懂「生成对抗网络 - GAN」基本原理+10种典型算法+13种应用 (easyai.tech)
2、代码实现集合:
GitHub - eriklindernoren/Keras-GAN: Keras implementations of Generative Adversarial Networks.
3、这里简单说以下
GNN的通俗理解基于两个对手之间相互博弈,共同进步。类似于:假设一个城市治安混乱,很快,这个城市里就会出现无数的小偷。在这些小偷中,有的可能是盗窃高手,有的可能毫无技术可言。假如这个城市开始整饬其治安,突然开展一场打击犯罪的「运动」,警察们开始恢复城市中的巡逻,很快,一批「学艺不精」的小偷就被捉住了。之所以捉住的是那些没有技术含量的小偷,是因为警察们的技术也不行了,在捉住一批低端小偷后,城市的治安水平变得怎样倒还不好说,但很明显,城市里小偷们的平均水平已经大大提高了。
在图像处理方面可以这么理解:你用真实图片和生成器(Generater)生成的虚假图片共同训练判别器(Discriminator),以致于其能够达到区分真假的功能。生成器(Generater)利用你随机输入的数字生成其对应与真实图片类似的图片,反复训练以至于能够生成越来越逼真的图片。
其图片如下:
生成图片和训练代码用的是卷积。
4、其损失函数
其实对于两者的损失函数可以分开考虑。对于判别器(Discriminator),就是其真实图片损失函数和制造的假图片损失函数求和。valid、fake对应的是其虚假标签;imgs,gen_imgs分别是真实图片和虚假图片。
d_loss_real = self.discriminator.train_on_batch(imgs, valid)d_loss_fake = self.discriminator.train_on_batch(gen_imgs, fake)
对于生成器(Generater),其先将判别器(Discriminator)设置为预测状态,然后X将传入生成器(Generater),将生成的虚假图片输入判别器(Discriminator)进行判别,这样就可以达到更新生成器(Generater)的参数。这里0(真)、1(假)主要是当Discriminator识别效果很好,说明Generater需要努力学习,才能继续蒙混过关,所以其损失值大;反之亦然。
其代码实现时将其标签直接传为了1(真实图片):
5、其代码实现:
代码时相对于Mnist实现,通过生成Mnist图像蒙混判别器(Discriminator)。
from __future__ import print_function, divisionimport tensorflow as tfimport osos.environ["CUDA_VISIBLE_DEVICES"] = "0"config = tf.ConfigProto(allow_soft_placement = True)gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction = 0.5)config.gpu_options.allow_growth = Truesess0 = tf.InteractiveSession(config = config)from keras.datasets import mnistfrom keras.layers import Input, Dense, Reshape, Flatten, Dropoutfrom keras.layers import BatchNormalization, Activation, ZeroPadding2Dfrom keras.layers.advanced_activations import LeakyReLUfrom keras.layers.convolutional import UpSampling2D, Conv2Dfrom keras.models import Sequential, Modelfrom keras.optimizers import Adamimport matplotlib.pyplot as pltimport sysimport numpy as npclass GAN():def __init__(self):self.img_rows = 28self.img_cols = 28self.channels = 1self.img_shape = (self.img_rows, self.img_cols, self.channels)self.latent_dim = 100optimizer = Adam(0.0002, 0.5)# Build and compile the discriminatorself.discriminator = self.build_discriminator()pile(loss='binary_crossentropy',optimizer=optimizer,metrics=['accuracy'])# Build the generatorself.generator = self.build_generator()# The generator takes noise as input and generates imgsz = Input(shape=(self.latent_dim,))img = self.generator(z)# For the combined model we will only train the generatorself.discriminator.trainable = False# The discriminator takes generated images as input and determines validityvalidity = self.discriminator(img)# The combined model (stacked generator and discriminator)# Trains the generator to fool the bined = Model(z, validity)pile(loss='binary_crossentropy', optimizer=optimizer)def build_generator(self):model = Sequential()model.add(Dense(256, input_dim=self.latent_dim))model.add(LeakyReLU(alpha=0.2))model.add(BatchNormalization(momentum=0.8))model.add(Dense(512))model.add(LeakyReLU(alpha=0.2))model.add(BatchNormalization(momentum=0.8))model.add(Dense(1024))model.add(LeakyReLU(alpha=0.2))model.add(BatchNormalization(momentum=0.8))model.add(Dense(np.prod(self.img_shape), activation='tanh'))model.add(Reshape(self.img_shape))model.summary()noise = Input(shape=(self.latent_dim,))img = model(noise)return Model(noise, img)def build_discriminator(self):model = Sequential()model.add(Flatten(input_shape=self.img_shape))model.add(Dense(512))model.add(LeakyReLU(alpha=0.2))model.add(Dense(256))model.add(LeakyReLU(alpha=0.2))model.add(Dense(1, activation='sigmoid'))model.summary()img = Input(shape=self.img_shape)validity = model(img)return Model(img, validity)def train(self, epochs, batch_size=128, sample_interval=50):# Load the dataset(X_train, _), (_, _) = mnist.load_data()# Rescale -1 to 1X_train = X_train / 127.5 - 1.X_train = np.expand_dims(X_train, axis=3)# Adversarial ground truthsvalid = np.ones((batch_size, 1))fake = np.zeros((batch_size, 1))for epoch in range(epochs):# ---------------------# Train Discriminator# ---------------------# Select a random batch of imagesidx = np.random.randint(0, X_train.shape[0], batch_size)imgs = X_train[idx]noise = np.random.normal(0, 1, (batch_size, self.latent_dim))# Generate a batch of new imagesgen_imgs = self.generator.predict(noise)# Train the discriminatord_loss_real = self.discriminator.train_on_batch(imgs, valid)d_loss_fake = self.discriminator.train_on_batch(gen_imgs, fake)d_loss = 0.5 * np.add(d_loss_real, d_loss_fake)# ---------------------# Train Generator# ---------------------noise = np.random.normal(0, 1, (batch_size, self.latent_dim))# Train the generator (to have the discriminator label samples as valid)g_loss = bined.train_on_batch(noise, valid)# Plot the progressprint ("%d [D loss: %f, acc.: %.2f%%] [G loss: %f]" % (epoch, d_loss[0], 100*d_loss[1], g_loss))# If at save interval => save generated image samplesif epoch % sample_interval == 0:self.sample_images(epoch)def sample_images(self, epoch):r, c = 5, 5noise = np.random.normal(0, 1, (r * c, self.latent_dim))gen_imgs = self.generator.predict(noise)# Rescale images 0 - 1gen_imgs = 0.5 * gen_imgs + 0.5fig, axs = plt.subplots(r, c)cnt = 0for i in range(r):for j in range(c):axs[i,j].imshow(gen_imgs[cnt, :,:,0], cmap='gray')axs[i,j].axis('off')cnt += 1fig.savefig("images/%d.png" % epoch)plt.close()if __name__ == '__main__':gan = GAN()gan.train(epochs=30000, batch_size=1, sample_interval=200)
其结果如下(训练时间不会很久,有兴趣的可以试试):
通过结果可以发现 随着训练步数的增加,生成的图片越来越能以假乱真。