Center-based 3D Object Detection and Tracking
3D Object Detection and Tracking using center points in the bird-eye view.
Center-based 3D Object Detection and Tracking,
Tianwei Yin, Xingyi Zhou, Philipp Krähenbühl,
arXiv technical report (arXiv 2006.11275)
@article{yin2021center,
title={Center-based 3D Object Detection and Tracking},
author={Yin, Tianwei and Zhou, Xingyi and Kr{\"a}henb{\"u}hl, Philipp},
journal={CVPR},
year={2021},
}
NEWS
[2021-06-20] The real time version of CenterPoint ranked 2nd in the Waymo Real-time 3D detection challenge (72.8 mAPH / 57.1 ms). The corresponding techical report is available at URL. Please stay tuned for the updates.
[2021-04-13] Better nuScenes results by fixing sync-bn bug and using stronger augmentations. Plese refer to nuScenes.
[2021-02-28] CenterPoint is accepted at CVPR 2021 🔥
[2021-01-06] CenterPoint v0.1 is released. Without bells and whistles, we rank first among all Lidar-only methods on Waymo Open Dataset with a single model that runs at 11 FPS. Check out CenterPoint's model zoo for Waymo and nuScenes.
[2020-12-11] 3 out of the top 4 entries in the recent NeurIPS 2020 nuScenes 3D Detection challenge used CenterPoint. Congratualations to other participants and please stay tuned for more updates on nuScenes and Waymo soon.
Contact
Any questions or suggestions are welcome!
Tianwei Yin yintianwei@utexas.edu Xingyi Zhou zhouxy@cs.utexas.edu
Abstract
Three-dimensional objects are commonly represented as 3D boxes in a point-cloud. This representation mimics the well-studied image-based 2D bounding-box detection but comes with additional challenges. Objects in a 3D world do not follow any particular orientation, and box-based detectors have difficulties enumerating all orientations or fitting an axis-aligned bounding box to rotated objects. In this paper, we instead propose to represent, detect, and track 3D objects as points. Our framework, CenterPoint, first detects centers of objects using a keypoint detector and regresses to other attributes, including 3D size, 3D orientation, and velocity. In a second stage, it refines these estimates using additional point features on the object. In CenterPoint, 3D object tracking simplifies to greedy closest-point matching. The resulting detection and tracking algorithm is simple, efficient, and effective. CenterPoint achieved state-of-the-art performance on the nuScenes benchmark for both 3D detection and tracking, with 65.5 NDS and 63.8 AMOTA for a single model. On the Waymo Open Dataset, CenterPoint outperforms all previous single model method by a large margin and ranks first among all Lidar-only submissions.
Highlights
-
Simple: Two sentences method summary: We use standard 3D point cloud encoder with a few convolutional layers in the head to produce a bird-eye-view heatmap and other dense regression outputs including the offset to centers in the previous frame. Detection is a simple local peak extraction with refinement, and tracking is a closest-distance matching.
-
Fast and Accurate: Our best single model achieves 71.9 mAPH on Waymo and 65.5 NDS on nuScenes while running at 11FPS+.
-
Extensible: Simple replacement for anchor-based detector in your novel algorithms.
Main results
3D detection on Waymo test set
#Frame | Veh_L2 | Ped_L2 | Cyc_L2 | MAPH | FPS | |
---|---|---|---|---|---|---|
VoxelNet | 1 | 71.9 | 67.0 | 68.2 | 69.0 | 13 |
VoxelNet | 2 | 73.0 | 71.5 | 71.3 | 71.9 | 11 |
3D detection on Waymo domain adaptation test set
#Frame | Veh_L2 | Ped_L2 | Cyc_L2 | MAPH | FPS | |
---|---|---|---|---|---|---|
VoxelNet | 2 | 56.1 | 47.8 | 65.2 | 56.3 | 11 |
3D detection on nuScenes test set
MAP ↑ | NDS ↑ | PKL ↓ | FPS ↑ | |
---|---|---|---|---|
VoxelNet | 58.0 | 65.5 | 0.69 | 11 |
3D tracking on Waymo test set
#Frame | Veh_L2 | Ped_L2 | Cyc_L2 | MOTA | FPS | |
---|---|---|---|---|---|---|
VoxelNet | 2 | 59.4 | 56.6 | 60.0 | 58.7 | 11 |
3D Tracking on nuScenes test set
AMOTA ↑ | AMOTP ↓ | |
---|---|---|
VoxelNet (flip test) | 63.8 | 0.555 |
All results are tested on a Titan RTX GPU with batch size 1.
Third-party resources
- ONCE_Benchmark: Implementation of CenterPoint on the ONCE dataset
- CenterPoint-KITTI: Reimplementation of CenterPoint on the KITTI dataset
- AFDet: another work inspired by CenterNet achieves good performance on KITTI/Waymo dataset.
- mmdetection3d: CenterPoint in mmdet framework.
Use CenterPoint
Installation
Please refer to INSTALL to set up libraries needed for distributed training and sparse convolution.
Benchmark Evaluation and Training
Please refer to GETTING_START to prepare the data. Then follow the instruction there to reproduce our detection and tracking results. All detection configurations are included in configs.
Develop
If you are interested in training CenterPoint on a new dataset, use CenterPoint in a new task, or use a new network architecture for CenterPoint, please refer to DEVELOP. Feel free to send us an email for discussions or suggestions.
ToDo List
- Support visualization with Open3D
- Colab demo
- Docker
License
CenterPoint is release under MIT license (see LICENSE). It is developed based on a forked version of det3d. We also incorperate a large amount of code from CenterNet and CenterTrack. See the NOTICE for details. Note that both nuScenes and Waymo datasets are under non-commercial licenses.
Acknowlegement
This project is not possible without multiple great opensourced codebases. We list some notable examples below.