https://github.com/bwconrad/stylegan-latent-embedding

Embedding images into StyleGAN's latent space with high reconstruction quality.
https://github.com/bwconrad/stylegan-latent-embedding

Last synced: about 1 month ago
JSON representation

Embedding images into StyleGAN's latent space with high reconstruction quality.

• Host: GitHub
• URL: https://github.com/bwconrad/stylegan-latent-embedding
• Owner: bwconrad
• Created: 2020-07-12T20:06:01.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2024-01-23T18:41:36.000Z (12 months ago)
• Last Synced: 2024-05-14T00:18:10.723Z (8 months ago)
• Language: Jupyter Notebook
• Homepage:
• Size: 93.6 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# StyleGAN Latent Embedding



  

  

  



Embed any image into StyleGAN's latent space. Based off of [PULSE's](https://github.com/adamian98/pulse) optimization method and codebase.

## How To Run

### Requirements

Run ```pip install -r requirements.txt``` to install all required packages.

### Usage

Run ```python run.py -input_dir ``` to embed and save all images in the input

directory.

- ```-save_intermediate```: save the current image after each optimization step.

- ```-save_latent```: save the final latent and noise vector to file which can be used by

```decode.py```.

```embed.ipynb``` offers an interface to embed single images and visualize the reconstruction error.

Run ```python decode.py -input ``` to reconstruct the image of a given saved latent and

noise vector from using the ```-save_latent``` flag with ```run.py```.