It's from same GAN though, C(x_r) and C(x_f) come from the same neural network. It's how realistic real data is compared to fake data (and vice-versa) as determ

It's from same GAN though, C(x_r) and C(x_f) come from the same neural network. It's how realistic real data is compared to fake data (and vice-versa) as determined by C (discriminator without activation function).