https://github.com/timsainb/gaia

Generative Adversarial Interpolative Autoencoder (GAIA) is a Generative Adversarial Network (GAN) made up of Autoencoders (AE) trained explicitly on interpolations to promote convexity and better latent interpolations.
https://github.com/timsainb/gaia

Last synced: 3 months ago
JSON representation

Generative Adversarial Interpolative Autoencoder (GAIA) is a Generative Adversarial Network (GAN) made up of Autoencoders (AE) trained explicitly on interpolations to promote convexity and better latent interpolations.

• Host: GitHub
• URL: https://github.com/timsainb/gaia
• Owner: timsainb
• Created: 2018-07-16T00:41:58.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2019-09-05T18:20:35.000Z (over 5 years ago)
• Last Synced: 2024-10-14T15:42:05.115Z (3 months ago)
• Language: Jupyter Notebook
• Size: 18.4 MB
• Stars: 77
• Watchers: 7
• Forks: 19
• Open Issues: 1
• Metadata Files:
  • Readme: readme.md

Awesome Lists containing this project

README

        
### Generative Adversarial Interpolative Autoencoding (GAIA)

#### Authors: Tim Sainburg, Marvin Thielk, Brad Theilman, Benjamin Migliori, Tim Gentner (UCSD)

The **Generative Adversarial Interpolative Autoencoder (GAIA; [Paper](http://timsainburg.com/assets/PDF/GAIA.pdf); [Blog post](http://timsainburg.com/gaia.html#gaia))** is novel hybrid between the Generative Adversarial Network (GAN) and the Autoencoder (AE). The purpose of GAIA is to address three issues which exist in GANs and AEs:

1. GANs are not bidirectional

2. Autoencoders produce blurry images

3. Autoencoder latent spaces are not convex

![Morph Image](images/celeb-morph.png)

#### Instructions

1. Download the GAIA dataset with the notebook '*Create_CELEBA-HQ.ipynb*'

1. Download the trained weights '*download_weights.ipynb*'

2. Run the notebook '*GAIA simple example.ipynb*'

**Note**: I'm currently in the process of rewriting this code to be cleaner, include more features, etc. For now, this is just the version of the code used in the Arxiv paper.

![Morph Image](images/celeb-attrs.png)

References  

- Multimodal Unsupervised Image-to-Image Translation ([Paper](https://arxiv.org/abs/1804.04732); [Author implementation](https://github.com/NVlabs/MUNIT); [Tensorflow implementation](https://github.com/taki0112/MUNIT-Tensorflow))

- [BEGAN](https://arxiv.org/abs/1703.10717)

- [VAEGAN](https://arxiv.org/abs/1512.09300)

- Progressively Growing GANs ([Paper](https://arxiv.org/abs/1710.10196), [Code](https://github.com/tkarras/progressive_growing_of_gans))

```python

```