• Anaconda or Miniconda
• Python = 3.8
• PyTorch = 1.10.2
• NVIDIA GPU (24 GB recommended) + CUDA = 11.3
• Linux

1. Clone repo

   git clone https://github.com/95anantsingh/NYU-SuperGAN.git
   cd NYU-SuperGAN

2. Create conda environment

   conda env create -f environment.yml

3. Download weights

   wget -i SuperGAN/weight_urls --directory-prefix SuperGAN/data/weights

cd SuperGAN
conda activate SuperGAN
python faceswap.py

Now play data/output/output.mp4 to see results
Make necessary changes in faceswap.py for different input videos.

1. SuperGAN contains all the project files
2. SuperGAN/data contains input output videos and pretrained weights
3. SuperGAN/faceswap.py main inference file

Project report can be found at docs/SuperGAN_Report.pdf

Ideal setup is with NVIDIA Quadro RTX 8000 (48 GB GDDR6) on Linux

SuperGAN is released under Apache License Version 2.0.

If you have any question, please email anant.singh@nyu.edu

This Project was part of graduate level Deep Learning course at New York University