/ss-ood

Self-Supervised Learning for OOD Detection (NeurIPS 2019)

Primary LanguagePythonMIT LicenseMIT

Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty

This repository contains the dataset and some code for the paper Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty by Dan Hendrycks, Mantas Mazeika, Saurav Kadavath, and Dawn Song.

We show that self-supervised learning can tremendously improve out-of-distribution detection as well as various types of robustness.

Download the one class ImageNet test set here. The one class ImageNet training set is here.

The code requires PyTorch 1.0 + and Python 3+.

Citation

If you find this useful in your research, please consider citing:

@article{hendrycks2019selfsupervised,
  title={Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty},
  author={Dan Hendrycks and Mantas Mazeika and Saurav Kadavath and Dawn Song},
  journal={Advances in Neural Information Processing Systems (NeurIPS)},
  year={2019}
}