James TompGAN - on GitHub

May 2020

TensorFlow implementation of NVIDIA's GauGAN for semantic image synthesis, made with Steven Cheung (cheungs01), Jason Ho (chekfung), and Jeremy Chen (jchen20) as a final project for Brown CS 143 - Computer Vision, taught by Professor James Tompkin.

We trained and tested the project on the MIT ADE20K dataset. We plan to publish a Python package (lemonADE20K) to make querying and manipulating the dataset easier, since the provided index for ADE20K is in MATLAB.

Our project report is available here.