LPF-Defense

The Code and Data for the paper "LPF-Defense: 3D Adversarial Defense based on Frequency Analysis", by Hanieh Naderi, Kimia Noorbakhsh*, Arian Etemadi*, and Shohreh Kasaei.

Prerequisites

This code depends on the following packages:

1. pyshtools
2. PyTorch
3. torchvision
4. numpy
5. matplotlib

Code Structure

The instructions on running each part of the code is explained in the README file of each folder separately.

Data

The ModelNet40 data we used in our paper is uploaded in folder model/Data/. The scanobjectnn Dataset and the subset of ShapeNet Dataset used in the paper are available at https://hkust-vgd.github.io/scanobjectnn/ and https://github.com/thuml/Metasets.

Citation

If you use parts of the code in this repository for your own research, please consider citing:

@article{10.1371/journal.pone.0271388,
    doi = {10.1371/journal.pone.0271388},
    author = {Naderi, Hanieh AND Noorbakhsh, Kimia AND Etemadi, Arian AND Kasaei, Shohreh},
    journal = {PLOS ONE},
    publisher = {Public Library of Science},
    title = {LPF-Defense: 3D adversarial defense based on frequency analysis},
    year = {2023},
    month = {02},
    volume = {18},
    url = {https://doi.org/10.1371/journal.pone.0271388},
    pages = {1-19},
    number = {2},
}