PGGAN Pytorch Implement

One day, I got tired of debugging those discriminative models against accuracy, F1-score, and other metrics. I wanted to have some fun. So I implemented the vanilla DCGAN, which generated fuzzy results. I wasn’t very happy with it, so I found PGGAN, but it only had tensorflow version of the code, and the PGGAN itself was a huge complicatedness. I spent two weeks reproducing the Pytorch version PGGAN and used it to generate to 128×128 size images. Then my GPU resources couldn’t support. But the whole process helped me practice a lot in object-oriented programming and building larger pipeline.