APDrawingGAN

Why we included this project

APDrawingGAN is a research implementation of a hierarchical GAN that converts a face photo into an artistic portrait drawing. Instead of relying on one global generator, the model breaks the face into regions such as eyes, nose, mouth, and hair, and trains a separate generator and discriminator for each, an architecture worth studying if you are building any part-based image synthesis pipeline. The repo includes pretrained weights, a preprocessing pipeline that aligns faces and extracts landmarks and background masks, and train/test scripts, so you can reproduce the paper's results or adapt the code to your own data. It is research code from a CVPR 2019 paper rather than a maintained product, so expect Python 2.7-era dependencies and some integration work on your side. For anyone prototyping artistic portrait generation or hierarchical GANs, it is still a solid starting point.