原创文章,转载请注明: 转载自慢慢的回味
本文链接地址: WGAN与DCGAN的区别
WGAN与DCGAN的区别
1 它们的主要区别在于损失函数不一样,DCGAN使用的是二位交叉熵(binary_crossentropy),WGAN使用的是推土机距离(wasserstein_loss),即目标值与预测值乘积的均值,所以鉴别网络末端不再使用激活函数sigmoid,而是直接输出全连接网络的值。
def wasserstein_loss(self, y_true, y_pred): return K.mean(y_true * y_pred) |
这儿需要注意的是:
对于真实的图片,y_true即目标值是-1,参考如下代码。而预测值是0-1,所有预测为真时(鉴别成功),值为-1×1=-1,为伪造的图片时(鉴别失败),值为-1×0=0,即鉴别失败时损失大(0>-1);
对于伪造的图片,y_true即目标值是1,参考如下代码。而预测值是0-1,所有预测为真时(鉴别失败),值为1×1=1,为伪造的图片时(鉴别成功),值为1×0=0,即鉴别失败时损失大(1>0);
# Adversarial ground truths valid = -np.ones((batch_size, 1)) fake = np.ones((batch_size, 1)) |
具体代码对比参看:WGAN与DCGAN的代码对比
2 另一个区别是,WGAN在鉴别网络训练时需要进行剪裁(clip),以防止权值更新太快。
# Clip critic weights for l in self.critic.layers: weights = l.get_weights() weights = [np.clip(w, -self.clip_value, self.clip_value) for w in weights] l.set_weights(weights) |
附基于Keras的WGAN测试程序
from __future__ import print_function, division from keras.datasets import mnist from keras.layers import Input, Dense, Reshape, Flatten, Dropout from keras.layers import BatchNormalization, Activation, ZeroPadding2D from keras.layers.advanced_activations import LeakyReLU from keras.layers.convolutional import UpSampling2D, Conv2D from keras.models import Sequential, Model from keras.optimizers import RMSprop import keras.backend as K import matplotlib.pyplot as plt import sys import numpy as np class WGAN(): def __init__(self): self.img_rows = 28 self.img_cols = 28 self.channels = 1 self.img_shape = (self.img_rows, self.img_cols, self.channels) self.latent_dim = 100 # Following parameter and optimizer set as recommended in paper self.n_critic = 5 self.clip_value = 0.01 optimizer = RMSprop(lr=0.00005) # Build and compile the critic self.critic = self.build_critic() self.critic.compile(loss=self.wasserstein_loss, optimizer=optimizer, metrics=['accuracy']) # Build the generator self.generator = self.build_generator() # The generator takes noise as input and generated imgs z = Input(shape=(self.latent_dim,)) img = self.generator(z) # For the combined model we will only train the generator self.critic.trainable = False # The critic takes generated images as input and determines validity valid = self.critic(img) # The combined model (stacked generator and critic) self.combined = Model(z, valid) self.combined.compile(loss=self.wasserstein_loss, optimizer=optimizer, metrics=['accuracy']) def wasserstein_loss(self, y_true, y_pred): return K.mean(y_true * y_pred) def build_generator(self): model = Sequential() model.add(Dense(128 * 7 * 7, activation="relu", input_dim=self.latent_dim)) model.add(Reshape((7, 7, 128))) model.add(UpSampling2D()) model.add(Conv2D(128, kernel_size=4, padding="same")) model.add(BatchNormalization(momentum=0.8)) model.add(Activation("relu")) model.add(UpSampling2D()) model.add(Conv2D(64, kernel_size=4, padding="same")) model.add(BatchNormalization(momentum=0.8)) model.add(Activation("relu")) model.add(Conv2D(self.channels, kernel_size=4, padding="same")) model.add(Activation("tanh")) model.summary() noise = Input(shape=(self.latent_dim,)) img = model(noise) return Model(noise, img) def build_critic(self): model = Sequential() model.add(Conv2D(16, kernel_size=3, strides=2, input_shape=self.img_shape, padding="same")) model.add(LeakyReLU(alpha=0.2)) model.add(Dropout(0.25)) model.add(Conv2D(32, kernel_size=3, strides=2, padding="same")) model.add(ZeroPadding2D(padding=((0,1),(0,1)))) model.add(BatchNormalization(momentum=0.8)) model.add(LeakyReLU(alpha=0.2)) model.add(Dropout(0.25)) model.add(Conv2D(64, kernel_size=3, strides=2, padding="same")) model.add(BatchNormalization(momentum=0.8)) model.add(LeakyReLU(alpha=0.2)) model.add(Dropout(0.25)) model.add(Conv2D(128, kernel_size=3, strides=1, padding="same")) model.add(BatchNormalization(momentum=0.8)) model.add(LeakyReLU(alpha=0.2)) model.add(Dropout(0.25)) model.add(Flatten()) model.add(Dense(1)) 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 1 X_train = (X_train.astype(np.float32) - 127.5) / 127.5 X_train = np.expand_dims(X_train, axis=3) # Adversarial ground truths valid = -np.ones((batch_size, 1)) fake = np.ones((batch_size, 1)) for epoch in range(epochs): for _ in range(self.n_critic): # --------------------- # Train Discriminator # --------------------- # Select a random batch of images idx = np.random.randint(0, X_train.shape[0], batch_size) imgs = X_train[idx] # Sample noise as generator input noise = np.random.normal(0, 1, (batch_size, self.latent_dim)) # Generate a batch of new images gen_imgs = self.generator.predict(noise) # Train the critic d_loss_real = self.critic.train_on_batch(imgs, valid) d_loss_fake = self.critic.train_on_batch(gen_imgs, fake) d_loss = 0.5 * np.add(d_loss_fake, d_loss_real) # Clip critic weights for l in self.critic.layers: weights = l.get_weights() weights = [np.clip(w, -self.clip_value, self.clip_value) for w in weights] l.set_weights(weights) # --------------------- # Train Generator # --------------------- g_loss = self.combined.train_on_batch(noise, valid) # Plot the progress print ("%d [D loss: %f] [G loss: %f]" % (epoch, 1 - d_loss[0], 1 - g_loss[0])) # If at save interval => save generated image samples if epoch % sample_interval == 0: self.sample_images(epoch) def sample_images(self, epoch): r, c = 5, 5 noise = np.random.normal(0, 1, (r * c, self.latent_dim)) gen_imgs = self.generator.predict(noise) # Rescale images 0 - 1 gen_imgs = 0.5 * gen_imgs + 0.5 fig, axs = plt.subplots(r, c) cnt = 0 for 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 += 1 fig.savefig("images/mnist_%d.png" % epoch) plt.close() if __name__ == '__main__': wgan = WGAN() wgan.train(epochs=4000, batch_size=32, sample_interval=50) |
本作品采用知识共享署名 4.0 国际许可协议进行许可。