https://github.com/ffwff/ganimu

Neural networks that learn to generate anime faces.
https://github.com/ffwff/ganimu

Last synced: about 1 year ago
JSON representation

Neural networks that learn to generate anime faces.

• Host: GitHub
• URL: https://github.com/ffwff/ganimu
• Owner: ffwff
• Created: 2018-10-02T12:15:48.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2019-02-17T11:48:12.000Z (over 7 years ago)
• Last Synced: 2025-05-12T22:55:30.283Z (about 1 year ago)
• Language: Python
• Homepage:
• Size: 644 KB
• Stars: 10
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# ganimu

Generating anime faces with deep convolution neural networks.

## Generating with a GAN network

Montage of 120 faces generated by a GAN. The model was trained for 71 epochs and the results are pretty cute(-ish)

An interactive web demo lies [here](https://fw.neocities.org/ganime.html).

![GAN](montage-gan.jpg)

## Generating with a VAE network

Montage of 90 faces generated by a [VAEGAN](https://arxiv.org/pdf/1512.09300.pdf). It's a lot blurrier than GAN, possibly due to the high weight of the L2 reconstruction loss. Due to performance and memory constraints I wasn't able to test this for long, however, feel free to try tuning the hyperparameters.

![GAN](montage-vae.jpg)