https://github.com/jmiller656/DiscoGAN-Tensorflow

An implementation of DiscoGAN in tensorflow
https://github.com/jmiller656/DiscoGAN-Tensorflow

discogan generative-adversarial-networks machine-learning neural-networks tensorflow

Last synced: 3 months ago
JSON representation

An implementation of DiscoGAN in tensorflow

• Host: GitHub
• URL: https://github.com/jmiller656/DiscoGAN-Tensorflow
• Owner: jmiller656
• License: mit
• Created: 2017-03-29T04:35:56.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-06-26T00:29:42.000Z (over 7 years ago)
• Last Synced: 2024-04-27T23:59:20.015Z (7 months ago)
• Topics: discogan, generative-adversarial-networks, machine-learning, neural-networks, tensorflow
• Language: Python
• Size: 14.6 KB
• Stars: 79
• Watchers: 4
• Forks: 14
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# DiscoGAN for Tensorflow

An implementation of [Learning to Discover Cross-Domain Relations with Generative Adversarial Networks](https://arxiv.org/abs/1703.05192) written in tensorflow.

## Requirements

- Tensorflow 1.0.1

- scipy

## Training

`python main.py`

## Training details

Currently the data utils file works on domains from the celeba dataset

## Remarks

As it currently stands, I have refactored much of the model and extracted it to `discoGAN.py`. I will soon be making it take command line arguments, download datasets automatically, etc. As mentioned before, there are now some barebones utilities to work with the celeba dataset.