FrequencyLowCut Pooling

Code for FrequencyLowCut Pooling - Plug & Play against Catastrophic Overfitting and Fix your downsampling ASAP! Be natively more robust via Aliasing and Spectral Artifact free Pooling

We provide our FrequencyLowCut (FLC) module and our Aliasing and Sinc Artifact free Pooling (ASAP) as well as examples how to implement them into common CNN structures.

The code for adversarial training used in our paper can be found here.

Citation

Would you like to reference our FLC Pooling and ASAP?

Then consider citing our paper and paper:

@inproceedings{grabinski2022frequencylowcut,
  title     = {FrequencyLowCut Pooling--Plug \& Play against Catastrophic Overfitting},
  author    = {Grabinski, Julia and Jung, Steffen and Keuper, Janis and Keuper, Margret},
  booktitle = {European Conference on Computer Vision},
  year      = {2022},
  url       = {https://arxiv.org/abs/2204.00491}
}

@article{grabinski2023fix,
  title    = {Fix your downsampling ASAP! Be natively more robust via Aliasing and Spectral Artifact free Pooling},
  author   = {Grabinski, Julia and Keuper, Janis and Keuper, Margret},
  journal  = {arXiv preprint arXiv:2307.09804},
  year     = {2023}
}