Simple MNIST GAN using TensorflowJS

Hand-written digit generation using Generative Adversarial Network. A simple TensorflowJS implementation that clocks in at less than 100 lines of code.

Early stages:
Getting better:
Later still:

Click Train to train for (an additional) 5) epochs. Click Load weights to restore pre-trained weights for the Generator. Click Sample image to generate a sample output using the current weights. The network should start to converge after 15-20 epochs.



Loading resources...this may take a few seconds




Fork me on GitHub