Shape your Space: A Gaussian Mixture Regularization Approach to Deterministic Autoencoders PyTorch

PyTorch implementation of the NeurIPS 2021 paper "Shape your Space: A Gaussian Mixture Regularization Approach to Deterministic Autoencoders". The paper can be found here. The code allows the users to reproduce and extend the results reported in the paper. Please cite the above paper when reporting, reproducing or extending the results.

Purpose of the project

This software is a research prototype, solely developed for and published as part of the publication. It will neither be maintained nor monitored in any way.

Setup.

Create a conda virtual environment
Clone the repository
Activate the environment and run

cd GMM_DAE
pip install -r requirements.txt

Dataset

The provided implementation is tested on MNIST, FASHION MNIST, SVHN and CELEBA images.

For dataset preparation use the script generate_dataset.py in the scripts folder:

python scripts/generate_dataset.py <DATASETNAME> <DATASETPATH>

CelebA has a daily quota that can only be overcome by manually downloading the dataset and placing the following files in <DATASETPATH>/celeba:

img_align_celeba.zip
list_attr_celeba.txt
identity_CelebA.txt
list_bbox_celeba.txt
list_landmarks_align_celeba.txt
list_eval_partition.txt

The paths to the datasets need to be saved in the config.ini file.

Usage

To run the code clone the repository and then run

python train.py <DATASETNAME> eg: MNIST, FASHIONMNIST, SVHN or CELEB

For FID computation we used the github repo pytorch-fid

License