Generative Adversarial Networks Projects
上QQ阅读APP看书,第一时间看更新

Training a 3D-GAN

Training a 3D-GAN is similar to training a vanilla GAN. We first train the discriminator network on both the generated images and the real images but freeze the generator network. Then, we train the generator network but freeze the discriminator network. We repeat this process for a specified number of epochs. During one iteration, we train both of the networks in a sequence. Training a 3D-GAN is an end-to-end training process. Let's work on these steps one by one.