cvlab-columbia / GenInt

Generative Interventions for Causal Learning

Code from the paper Generative Interventions for Causal Learning.

If you use this code, please consider citing the paper as:

@InProceedings{Mao_2021_CVPR,
    author    = {Mao, Chengzhi and Cha, Augustine and Gupta, Amogh and Wang, Hao and Yang, Junfeng and Vondrick, Carl},
    title     = {Generative Interventions for Causal Learning},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2021},
    pages     = {3947-3956}
}

@article{mao2020generative,
      title={Generative Interventions for Causal Learning}, 
      author={Chengzhi Mao and Augustine Cha and Amogh Gupta and Hao Wang and Junfeng Yang and Carl Vondrick},
      year={2020},
      eprint={2012.12265},
      journal={arXiv},      
}

Motivating Example: colored MNIST

Please go to ColorMNIST for instructions

Data Preparation

Generate data for GenInt

Please go to IntervenedGAN folder and follow the instruction in README.md and SETUP.md

Generate data for Observational GAN

Please go to BigGAN folder and follow instruction in README.md

Training the ImageNet models (download pretrained models)

Please go to training folder and follow the instructions in README.md