Skip to content

26789564/cvae-celeba

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

cvae-celeba

an implementation of cvae that allows setting attributes for celeba and setting labels for mnist


Results


How to run

data directory tree:
data/
  celeba/
    selected_list_attr_celeba.txt/
    original_list_attr_celeba.txt/
    img_align_celeba/
        selected_images/
        original_images/

use functions in dataloader.py to select desired attributes (e.g. in this repo 10 were chosen) and corresponding images
then run train.py with appropriate arguments
when running inference, adjust array c to create image with desired attributes (change ~line 117 in train.py, currently c is random)
setting reduction parameter in BCE loss to "sum" rather than "mean" produces high loss but better reconstruction results due to emphasis on reconstruction rather than divergence of distribution


Credits

starting code with MNIST was taken from https://github.com/timbmg/VAE-CVAE-MNIST

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%