Missing Balanced Adversarial Focal Training (BAFT) Implementation

Eaphan / Robust3DOD

A Comprehensive Study of the Robustness for LiDAR-based 3D Object Detectors against Adversarial Attacks

https://arxiv.org/abs/2212.10230

Apache License 2.0

92 stars 4 forks source link

Missing Balanced Adversarial Focal Training (BAFT) Implementation #4

Open miladabd opened 2 months ago

miladabd commented 2 months ago

Dear authors,

Thank you for sharing the official implementation of your work. I'm particularly interested in reproducing the adversarial training results using your Balanced Adversarial Focal Training (BAFT) approach.

I've noticed that the BAFT implementation seems to be missing from the current repository. Would it be possible to update the repo to include this?

Thank you for your time and consideration.

Eaphan commented 2 months ago

The BAFT implementation for adversarial point perturbation is here: https://github.com/Eaphan/Robust3DOD/blob/main/tools/train_utils/train_utils_BAFT_perturbation_demo.py

Please change the hyper-parameters according to the type of detectors you use, such as "key" and "model_name". Then use the file to replace the origin "train_utils.py"