Home

Awesome

<img src="docs/trailab.png" align="right" width="20%">

PDV

PDV is LiDAR 3D object detection method. This repository is based off [OpenPCDet].

Point Density-Aware Voxels for LiDAR 3D Object Detection
Jordan S. K. Hu, Tianshu Kuai, Steven L. Waslander
[Paper]

PDV diagram

Overview

Changelog

[2022-03-07] PDV v0.1.0 is released.

Model Zoo

KITTI 3D Object Detection Baselines

Selected supported methods are shown in the below table. The results are the 3D detection performance of moderate difficulty on the val set of KITTI dataset.

training timeCar@R40Pedestrian@R40Cyclist@R40logdownload
PDV~23 hours85.0557.4175.95logmodel-147M

Waymo Open Dataset Baselines

We could not provide the above pretrained models due to Waymo Dataset License Agreement, but you could easily achieve similar performance by training with the default configs. PDV is trained with 10% data (~16k frames) on 4 NVIDIA Tesla V100s GPUs.

Installation

Please refer to INSTALL.md for the installation of PDV.

Quick Demo

Please refer to DEMO.md for a quick demo to test with a pretrained model and visualize the predicted results on your custom data or the original KITTI data.

Getting Started

Please refer to GETTING_STARTED.md to learn more usage about this project.

License

PDV is released under the Apache 2.0 license.

Acknowledgement

We would like to thank the authors of OpenPCDet for their open source release of their codebase.

Citation

If you find this project useful in your research, please consider citing:

@article{PDV,
    title={Point Density-Aware Voxels for LiDAR 3D Object Detection},
    author={Jordan S. K. Hu and
            Tianshu Kuai and
            Steven L. Waslander},
    journal={CVPR},
    year={2022}
}