In 03_CGAN_MNIST
(1) d_loss_real = discriminator.train_on_batch(x=[X_batch, real_labels], y=real * (1 - smooth))
(2) d_loss_fake = discriminator.train_on_batch(x=[X_fake, random_labels], y=fake)
(3) d_g_loss_batch = d_g.train_on_batch(x=[z, random_labels], y=real)

To train the discriminator, you first train for X_batch with real_labels then X_fake with random_labels. I think this should be real_labels instead of random_labels in equation 2.

To train the generator, in equation (3), why you take random_lables for training the d_g instead of real_labels?
Thank you.

in ccGAN, we feed the img_real and img_fake to train the weights of discriminator. After that, we should train the generator weights. I think we should use the [masked_imgs, real] as the input and output pairs to train the generator, but you use the img_real to train that. When the input is img_real, how could generator learning anything to generate a new pic? Could you explain the reason? thanks.

for e in range(epochs + 1):
    for i in range(len(X_train) // batch_size):
        
        # Train Discriminator weights
        discriminator.trainable = True
        
        # Real samples
        img_real = X_train[i*batch_size:(i+1)*batch_size]
        real_labels = y_train[i*batch_size:(i+1)*batch_size]
        
        d_loss_real = discriminator.train_on_batch(x=img_real, y=[real, real_labels])
        
        # Fake Samples
        masked_imgs = mask_randomly(img_real)
        gen_imgs = generator.predict(masked_imgs)
        
        d_loss_fake = discriminator.train_on_batch(x=gen_imgs, y=[fake, fake_labels])
         
        # Discriminator loss
        d_loss_batch = 0.5 * (d_loss_real[0] + d_loss_fake[0])
        
        # Train Generator weights
        discriminator.trainable = False

        d_g_loss_batch = d_g.train_on_batch(x=img_real, y=real)     # =============> d_g_loss_batch = d_g.train_on_batch(x=masked_imgs, y=real) 

After training for 200 epochs with the CIFAR 10 dataset, I found that the DCGAN and CGAN generated images do not have enough variety. Most likely a mode collapse happening.

Hi there,

I have a question related to your awesomework.
Base on LSGAN paper, the loss function may looks like this on tensorflow version
D_loss = 0.5 * (tf.reduce_mean((D_real - 1)2) + tf.reduce_mean(D_fake2))
G_loss = 0.5 * tf.reduce_mean((D_fake - 1)**2)

I check your loss function which is MSE. Where do you define the a, b, c for LSGAN.
d_g.compile(optimizer=optimizer, loss='mse', metrics=['binary_accuracy'])

Thank you,

for e in range(epochs):
    for i in range(len(X_train) // batch_size):

        # Train Discriminator weights
        discriminator.trainable = True

        # Real samples
        X_batch = X_train[i*batch_size:(i+1)*batch_size]
        d_loss_real = discriminator.train_on_batch(x=X_batch, y=real * (1 - smooth))

        # Fake Samples
        z = np.random.normal(loc=0, scale=1, size=(batch_size, latent_dim))
        X_fake = generator.predict_on_batch(z)
        d_loss_fake = discriminator.train_on_batch(x=X_fake, y=fake)

        # Discriminator loss
        d_loss_batch = 0.5 * (d_loss_real[0] + d_loss_fake[0])

        # Train Generator weights
        discriminator.trainable = False   <<============
        d_g_loss_batch = d_g.train_on_batch(x=z, y=real)

Hello, I know the process of GAN but I am a little bit confused with the use of discriminator.trainable = False (see <<============) because in d_g, the discriminator is already false so can we remove this from here (<<============) or it has some other meaning here and I am not able to get it?

Please help me out. Thank you.