Skip to content

Code for Robustness-via-Synthesis: Robust training with generative adversarial perturbations

License

Notifications You must be signed in to change notification settings

ALLab-Boun/robustness-via-synthesis

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Robustness-via-Synthesis: Robust training with generative adversarial perturbations

Introduction

This paper presents a robust training algorithm where the adversarial perturbations are automatically synthesized from a random vector using a generator network. The classifier is trained with cross-entropy loss regularized with the optimal transport distance between the representations of the natural and synthesized adversarial samples. The proposed approach attains comparable robustness with various gradient-based and generative robust training techniques on CIFAR10, CIFAR100, SVHN, and Tiny ImageNet datasets. Code for CIFAR10 is provided in this repository. The codebase is modified from MadryLab's cifar10_challenge. Pretrained models for CIFAR10 and CIFAR100 are also shared.

Usage

For training

python train.py

For evaluation:

python eval_pgd_attack.py

Data and Models

cifar10_data folder and trained models for CIFAR10 and CIFAR100 can be accessed through the following link.

https://drive.google.com/drive/folders/1F6hHz1WbymE6w2hzWf6nNqMbd9-UvS5R?usp=share_link

Models for SVHN and TinyImagenet can be shared upon request.

Cite

If you find this work is useful, please cite the following:

@article{baytacs2022robustness,
  title={Robustness-via-synthesis: Robust training with generative adversarial perturbations},
  author={Bayta{\c{s}}, {\.I}nci M and Deb, Debayan},
  journal={Neurocomputing},
  year={2022},
  publisher={Elsevier}
}

Contact

inci.baytas@boun.edu.tr

About

Code for Robustness-via-Synthesis: Robust training with generative adversarial perturbations

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages