Instance Segmentation of LiDAR Point Clouds

Feihu Zhang; Chenye Guan; Jin Fang; Song Bai; Ruigang Yang; Philip H.S. Torr; Victor Prisacariu

doi:10.1109/ICRA40945.2020.9196622

Instance Segmentation of LiDAR Point Clouds

Feihu Zhang, Chenye Guan, Jin Fang, Song Bai, Ruigang Yang, Philip H.S. Torr, Victor Prisacariu

Computer Science

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

56 Scopus citations

Abstract

We propose a robust baseline method for instance segmentation which are specially designed for large-scale outdoor LiDAR point clouds. Our method includes a novel dense feature encoding technique, allowing the localization and segmentation of small, far-away objects, a simple but effective solution for single-shot instance prediction and effective strategies for handling severe class imbalances. Since there is no public dataset for the study of LiDAR instance segmentation, we also build a new publicly available LiDAR point cloud dataset to include both precise 3D bounding box and point-wise labels for instance segmentation, while still being about 3∼20 times as large as other existing LiDAR datasets. The dataset will be published at https://github.com/feihuzhang/LiDARSeg.

Original language	English
Title of host publication	2020 IEEE International Conference on Robotics and Automation, ICRA 2020
Pages	9448-9455
Number of pages	8
ISBN (Electronic)	9781728173955
DOIs	https://doi.org/10.1109/ICRA40945.2020.9196622
State	Published - May 2020
Event	2020 IEEE International Conference on Robotics and Automation, ICRA 2020 - Paris, France Duration: May 31 2020 → Aug 31 2020

Publication series

Name	Proceedings - IEEE International Conference on Robotics and Automation
ISSN (Print)	1050-4729

Conference

Conference	2020 IEEE International Conference on Robotics and Automation, ICRA 2020
Country/Territory	France
City	Paris
Period	5/31/20 → 8/31/20

Bibliographical note

Publisher Copyright:
© 2020 IEEE.

Funding

ACKNOWLEDGEMENT Research is mainly supported by Baidu’s Robotics and Auto-driving Lab, in part by the ERC grant ERC-2012-AdG 321162-HELIOS, EPSRC grant Seebibyte EP/M013774/1 and EPSRC/MURI grant EP/N019474/1. We would also like to acknowledge the Royal Academy of Engineering. Victor Adrian Prisacariu would like to thank the European Commission Project Multiple-actOrs Virtual Empathic CARegiver for the Elder (MoveCare).

Funders	Funder number
Multidisciplinary University Research Initiative	EP/N019474/1
Multidisciplinary University Research Initiative
Engineering and Physical Sciences Research Council	EP/M013774/1
Engineering and Physical Sciences Research Council

ASJC Scopus subject areas

Software
Control and Systems Engineering
Electrical and Electronic Engineering
Artificial Intelligence

Access to Document

10.1109/ICRA40945.2020.9196622

Cite this

@inproceedings{e155997350884279ba0bff2df247d703,

title = "Instance Segmentation of LiDAR Point Clouds",

abstract = "We propose a robust baseline method for instance segmentation which are specially designed for large-scale outdoor LiDAR point clouds. Our method includes a novel dense feature encoding technique, allowing the localization and segmentation of small, far-away objects, a simple but effective solution for single-shot instance prediction and effective strategies for handling severe class imbalances. Since there is no public dataset for the study of LiDAR instance segmentation, we also build a new publicly available LiDAR point cloud dataset to include both precise 3D bounding box and point-wise labels for instance segmentation, while still being about 3∼20 times as large as other existing LiDAR datasets. The dataset will be published at https://github.com/feihuzhang/LiDARSeg.",

author = "Feihu Zhang and Chenye Guan and Jin Fang and Song Bai and Ruigang Yang and Torr, \{Philip H.S.\} and Victor Prisacariu",

note = "Publisher Copyright: {\textcopyright} 2020 IEEE.; 2020 IEEE International Conference on Robotics and Automation, ICRA 2020 ; Conference date: 31-05-2020 Through 31-08-2020",

year = "2020",

month = may,

doi = "10.1109/ICRA40945.2020.9196622",

language = "English",

series = "Proceedings - IEEE International Conference on Robotics and Automation",

pages = "9448--9455",

booktitle = "2020 IEEE International Conference on Robotics and Automation, ICRA 2020",

}

Zhang, F, Guan, C, Fang, J, Bai, S, Yang, R, Torr, PHS & Prisacariu, V 2020, Instance Segmentation of LiDAR Point Clouds. in 2020 IEEE International Conference on Robotics and Automation, ICRA 2020., 9196622, Proceedings - IEEE International Conference on Robotics and Automation, pp. 9448-9455, 2020 IEEE International Conference on Robotics and Automation, ICRA 2020, Paris, France, 5/31/20. https://doi.org/10.1109/ICRA40945.2020.9196622

Instance Segmentation of LiDAR Point Clouds. / Zhang, Feihu; Guan, Chenye; Fang, Jin et al.
2020 IEEE International Conference on Robotics and Automation, ICRA 2020. 2020. p. 9448-9455 9196622 (Proceedings - IEEE International Conference on Robotics and Automation).

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

TY - GEN

T1 - Instance Segmentation of LiDAR Point Clouds

AU - Zhang, Feihu

AU - Guan, Chenye

AU - Fang, Jin

AU - Bai, Song

AU - Yang, Ruigang

AU - Torr, Philip H.S.

AU - Prisacariu, Victor

PY - 2020/5

Y1 - 2020/5

N2 - We propose a robust baseline method for instance segmentation which are specially designed for large-scale outdoor LiDAR point clouds. Our method includes a novel dense feature encoding technique, allowing the localization and segmentation of small, far-away objects, a simple but effective solution for single-shot instance prediction and effective strategies for handling severe class imbalances. Since there is no public dataset for the study of LiDAR instance segmentation, we also build a new publicly available LiDAR point cloud dataset to include both precise 3D bounding box and point-wise labels for instance segmentation, while still being about 3∼20 times as large as other existing LiDAR datasets. The dataset will be published at https://github.com/feihuzhang/LiDARSeg.

AB - We propose a robust baseline method for instance segmentation which are specially designed for large-scale outdoor LiDAR point clouds. Our method includes a novel dense feature encoding technique, allowing the localization and segmentation of small, far-away objects, a simple but effective solution for single-shot instance prediction and effective strategies for handling severe class imbalances. Since there is no public dataset for the study of LiDAR instance segmentation, we also build a new publicly available LiDAR point cloud dataset to include both precise 3D bounding box and point-wise labels for instance segmentation, while still being about 3∼20 times as large as other existing LiDAR datasets. The dataset will be published at https://github.com/feihuzhang/LiDARSeg.

UR - http://www.scopus.com/inward/record.url?scp=85089557480&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85089557480&partnerID=8YFLogxK

U2 - 10.1109/ICRA40945.2020.9196622

DO - 10.1109/ICRA40945.2020.9196622

M3 - Conference contribution

AN - SCOPUS:85089557480

T3 - Proceedings - IEEE International Conference on Robotics and Automation

SP - 9448

EP - 9455

BT - 2020 IEEE International Conference on Robotics and Automation, ICRA 2020

T2 - 2020 IEEE International Conference on Robotics and Automation, ICRA 2020

Y2 - 31 May 2020 through 31 August 2020

ER -

Instance Segmentation of LiDAR Point Clouds

Abstract

Publication series

Conference

Bibliographical note

Funding

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this