Please download the following data: point cloud, images, calibration files, labels, and re-organize the datasets as follows:
kitti
|- training
|- calib
|- image_2
|- label_2
|- velodyne
|- testing
|- calib
|- image_2
|- velodyne
Please follow the guidelines in PIXOR Implementation to install all the dependencies
To run our attack:
python run_attack.py --attack_type [perturb or add] --point_idx [point cloud index] --iter_num [number of iterations] --attack_lr [attack learning rate] --save_path [attack results save path]
If you find our work useful, please cite:
@InProceedings{Liu_2023_CVPR,
author = {Liu, Han and Wu, Yuhao and Yu, Zhiyuan and Vorobeychik, Yevgeniy and Zhang, Ning},
title = {SlowLiDAR: Increasing the Latency of LiDAR-Based Detection Using Adversarial Examples},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2023},
pages = {5146-5155}
}
Thanks for the open souce code https://github.com/philip-huang/PIXOR