/gan-sandbox

Play around with GANs

Primary LanguagePython

Play around with GANs

PyTorch implementation of Generative Adversarial Networks and play around with CelebA dataset, github-like Identicon images and Pokemon sprints.

Intermediate Results of Identicon Generation

Repository structure

├── bin
│   ├── train.py
│   ├── dashboard.ipynb
│   ├── session_1
│   ├── session_2
│   ├── ...
│   └── session_n
├── core
│   ├── dataloader
│   │   ├── transforms
│   │   └── utils
│   ├── engine
│   │   ├── config_file.py
│   │   ├── trainer.py
│   │   └── utils
│   └── models
│       ├── backbones
│       └── modules
├── data
├── docs
└── utils
  • data: contains actual data i.e. CelebA faces and Pokemon sprints, structured according to provider conventions
  • docs: any paper, notes, figures relevant to this repository
  • bin: dashboard.ipynb is our UI to setup experience sessions, setting up an associated directory session_i. The experiment is then launched through executing train.py, here for more details
  • core: contains definition of data processing and loading protocols, models and training engines

References

@incollection{NIPS2014_5423,
title = {Generative Adversarial Nets},
author = {Goodfellow, Ian and Pouget-Abadie, Jean and Mirza, Mehdi and Xu, Bing and Warde-Farley, David and Ozair, Sherjil and Courville, Aaron and Bengio, Yoshua},
booktitle = {Advances in Neural Information Processing Systems 27},
editor = {Z. Ghahramani and M. Welling and C. Cortes and N. D. Lawrence and K. Q. Weinberger},
pages = {2672--2680},
year = {2014},
publisher = {Curran Associates, Inc.},
url = {http://papers.nips.cc/paper/5423-generative-adversarial-nets.pdf}
}