Disclaimer: This is a part of a greater project where we experimented with generative models, you can find the complete project at repo
In this repo, I'll introduce my contribution to the project where I implemented pixelCNN and gatedPixelCNN models using pytorch. For pixelCNN I've implemented the following architecture:
And for the gatedPixelCNN, the main difference is replacing the Relu activation function with a gated function, the architecture mainly consists of gated-Blocks that consist of the following components:
PixelCNN samples:
GatedPixelCNN samples :
