Skip to content

Zhiyuan1991/proVLAE

Repository files navigation

proVLAE

Code for paper PROGRESSIVE LEARNING AND DISENTANGLEMENT OF HIERARCHICAL REPRESENTATIONS.

Requirements:

python 3.5
tensorflow 1.13.1 (also tested on 1.8.0, should be flexible)
pytorch 0.4.1 (for computing metrics only)
numpy
scipy 1.1.0 (You might encounter errors of "imsave" because "imsave is deprecated in SciPy 1.0.0, and will be removed in 1.2.")
moviepy (for generating .gif images)

Dataset

dSprite https://github.com/deepmind/dsprites-dataset
3DShapes https://github.com/deepmind/3d-shapes
celebA http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html (Align&Cropped Images)
please download those datasets and change the paths in "data_manager.py"

Scripts

train model: train_pro-VLAE.sh
view results: visualization.sh
compute metrics: compute_metric.sh

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published