Dual Contradistinctive Generative AutoEncoder (CVPR 2021)

Project | Paper | Usage | Citation

A generative autoencoder model with dual contradistinctive losses to improve generative autoencoder that performs simultaneous inference (reconstruction) and synthesis (sampling). Our model, named dual contradistinctive generative autoencoder (DC-VAE), integrates an instance-level discriminative loss (maintaining the instance-level fidelity for the reconstruction/synthesis) with a set-level adversarial loss (encouraging the set-level fidelity for there construction/synthesis), both being contradistinctive.

Usage

Clone the repository

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

Setup the conda environment

conda env create -f dcvae_env.yml
conda activate dcvae_env

Train the network on CIFAR-10
```
python train.py
```

Citation

If you find this project useful for your research, please cite the following work.

@InProceedings{Parmar_2021_CVPR,
    author    = {Parmar, Gaurav and Li, Dacheng and Lee, Kwonjoon and Tu, Zhuowen},
    title     = {Dual Contradistinctive Generative Autoencoder},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2021},
    pages     = {823-832}
}

Credits

We found the following libraries helpful in our research.

FID - computing the FID score
IS - computing the Inception Score.
AutoGAN - model architecture for the low resolution experiments experiments
ProGAN - model architecture for the high resolution experiments.

Acknowledgements

This work is funded by NSF IIS- 1717431 and NSF IIS-1618477. We thank Qualcomm Inc. for an award support. The work was performed when G. Parmar and D. Li were with UC San Diego.