Guided variational Autoencoder for Disentanglement Learning

Official Implementation for Guided variational Autoencoder for Disentanglement Learning (CVPR2020). This repo contains codes for the training and visualizing steps on MNIST and CelebA datasets in unsupervised way and supervised way respectively.

Installation

1. Clone this repo:

git clone https://github.com/mlpc-ucsd/Guided-VAE.git
cd Guided-VAE

2. Set up a new conda environment and activate it.

conda create -n GuidedVAE python=3.8
conda activate GuidedVAE

3. Install pytorch environment.

conda install pytorch torchvision cudatoolkit=11.1 -c pytorch -c nvidia
conda install tqdm pillow

Datasets Preparation

The structure of datasets should be

data_dir/
    celeba/
        images/
        list_attr_celeba.txt
    MNIST/
    	processed/
    	raw/

• For MNIST, torchvision will automatically download it.
• For CelebA, please refer to the CelebA Dataset page.

Training and Visualizing

• For MNIST

python main.py --dataset MNIST --dataroot /path/to/your/data_dir --output output_MNIST

• For CelebA

python main.py --dataset CelebA --dataroot /path/to/your/data_dir --output output_CelebA

For more options, please refer to main.py.

Citation

@inproceedings{ding2020guided,
  title={Guided variational autoencoder for disentanglement learning},
  author={Ding, Zheng and Xu, Yifan and Xu, Weijian and Parmar, Gaurav and Yang, Yang and Welling, Max and Tu, Zhuowen},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={7920--7929},
  year={2020}
}

License

This repository is released under the Apache License 2.0. License can be found in LICENSE file.