SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds

Meng Liu; Jianwei Niu; Yu Liu

doi:10.1109/RCAR58764.2023.10249561

SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds

Meng Liu^*
, Jianwei Niu
, Yu Liu

^*Corresponding author for this work

School of Computer Science and Engineering

China North Vehicle Research Institute

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

LiDAR 3D object detection for autonomous driving is an important issue. To address this issue, this paper provides a two-stage anchor-based solution. Firstly, voxel feature encoding and sparse convolution networks were employed for proposal generation. Secondly, we performed voxel feature aggregation and hierarchical feature learning in combination with the proposals, multi-level and multiscale voxel features, point encoding and voxel-wise ROI pooling. The aggregated object-wise voxel features were utilized to refine the proposals. Extensive experiments were conducted on three challenging datasets: KITTI, nuScenes and Waymo. Experimental results demonstrated that our SP-Net can effectively achieve multi-category 3D object detection in diverse scenarios offering satisfactory accuracy and speed (28.6 FPS).

Original language	English
Title of host publication	Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	358-365
Number of pages	8
ISBN (Electronic)	9798350327182
DOIs	https://doi.org/10.1109/RCAR58764.2023.10249561
State	Published - 2023
Event	2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023 - Datong, China Duration: 17 Jul 2023 → 20 Jul 2023

Publication series

Name	Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023

Conference

Conference	2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023
Country/Territory	China
City	Datong
Period	17/07/23 → 20/07/23

Access to Document

10.1109/RCAR58764.2023.10249561

Cite this

Liu, M., Niu, J., & Liu, Y. (2023). SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds. In Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023 (pp. 358-365). (Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/RCAR58764.2023.10249561

Liu, Meng ; Niu, Jianwei ; Liu, Yu. / SP-Net : A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds. Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023. Institute of Electrical and Electronics Engineers Inc., 2023. pp. 358-365 (Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023).

@inproceedings{bbe449a4c0bb4ae7af64130431f4637c,

title = "SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds",

abstract = "LiDAR 3D object detection for autonomous driving is an important issue. To address this issue, this paper provides a two-stage anchor-based solution. Firstly, voxel feature encoding and sparse convolution networks were employed for proposal generation. Secondly, we performed voxel feature aggregation and hierarchical feature learning in combination with the proposals, multi-level and multiscale voxel features, point encoding and voxel-wise ROI pooling. The aggregated object-wise voxel features were utilized to refine the proposals. Extensive experiments were conducted on three challenging datasets: KITTI, nuScenes and Waymo. Experimental results demonstrated that our SP-Net can effectively achieve multi-category 3D object detection in diverse scenarios offering satisfactory accuracy and speed (28.6 FPS).",

author = "Meng Liu and Jianwei Niu and Yu Liu",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023 ; Conference date: 17-07-2023 Through 20-07-2023",

year = "2023",

doi = "10.1109/RCAR58764.2023.10249561",

language = "英语",

series = "Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "358--365",

booktitle = "Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023",

address = "美国",

}

Liu, M, Niu, J & Liu, Y 2023, SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds. in Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023. Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023, Institute of Electrical and Electronics Engineers Inc., pp. 358-365, 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023, Datong, China, 17/07/23. https://doi.org/10.1109/RCAR58764.2023.10249561

SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds. / Liu, Meng; Niu, Jianwei ; Liu, Yu.
Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023. Institute of Electrical and Electronics Engineers Inc., 2023. p. 358-365 (Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - SP-Net

T2 - 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023

AU - Liu, Meng

AU - Niu, Jianwei

AU - Liu, Yu

PY - 2023

Y1 - 2023

N2 - LiDAR 3D object detection for autonomous driving is an important issue. To address this issue, this paper provides a two-stage anchor-based solution. Firstly, voxel feature encoding and sparse convolution networks were employed for proposal generation. Secondly, we performed voxel feature aggregation and hierarchical feature learning in combination with the proposals, multi-level and multiscale voxel features, point encoding and voxel-wise ROI pooling. The aggregated object-wise voxel features were utilized to refine the proposals. Extensive experiments were conducted on three challenging datasets: KITTI, nuScenes and Waymo. Experimental results demonstrated that our SP-Net can effectively achieve multi-category 3D object detection in diverse scenarios offering satisfactory accuracy and speed (28.6 FPS).

AB - LiDAR 3D object detection for autonomous driving is an important issue. To address this issue, this paper provides a two-stage anchor-based solution. Firstly, voxel feature encoding and sparse convolution networks were employed for proposal generation. Secondly, we performed voxel feature aggregation and hierarchical feature learning in combination with the proposals, multi-level and multiscale voxel features, point encoding and voxel-wise ROI pooling. The aggregated object-wise voxel features were utilized to refine the proposals. Extensive experiments were conducted on three challenging datasets: KITTI, nuScenes and Waymo. Experimental results demonstrated that our SP-Net can effectively achieve multi-category 3D object detection in diverse scenarios offering satisfactory accuracy and speed (28.6 FPS).

UR - https://www.scopus.com/pages/publications/85173623416

U2 - 10.1109/RCAR58764.2023.10249561

DO - 10.1109/RCAR58764.2023.10249561

M3 - 会议稿件

AN - SCOPUS:85173623416

T3 - Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023

SP - 358

EP - 365

BT - Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 17 July 2023 through 20 July 2023

ER -

Liu M, Niu J , Liu Y. SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds. In Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023. Institute of Electrical and Electronics Engineers Inc. 2023. p. 358-365. (Proceedings of the 2023 IEEE International Conference on Real-Time Computing and Robotics, RCAR 2023). doi: 10.1109/RCAR58764.2023.10249561

SP-Net: A Sparse Convolution and Point-Encoding Enhanced Network for 3D Object Detection in LiDAR Point Clouds

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this