Skip to content
Implementation of Post-Average adversarial defense
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Post-Average Adversarial Defense

Implementation of the Post-Average adversarial defense method as described in Bandlimiting Neural Networks Against Adversarial Attacks.

This implementation is based on PyTorch and uses the Foolbox toolbox to provide attacking methods.

robustml evaluation

This implementation supports the robustml API for evaluation.

To evaluate on CIFAR-10:

python <datasetPath>

To evaluate on ImageNet:

python <datasetPath>
You can’t perform that action at this time.