GAN: Generative Adversarial Networks

Evaluation

Loss

Discriminator Loss

Discriminator Accuracy

Generated images while training

The GAN was trained for 150 epochs: Each column represents a noise vector which is transformed by the generator into an image. There are ten different noise vectors. The i^th row represents the model's state at epoch i * 10.

Generated images

Let's generate 100 images after training the GAN i.e. after 150 epochs.