LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation

Zeyu Hu; Xuyang Bai; Runze Zhang; Xin Wang; Guangyuan Sun; Hongbo Fu; Chiew-Lan Tai

doi:10.1007/978-3-031-19812-0_15

LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation

Zeyu Hu^*, Xuyang Bai, Runze Zhang, Xin Wang, Guangyuan Sun, Hongbo Fu, Chiew-Lan Tai

^*Corresponding author for this work

School of Creative Media

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

16 Citations (Scopus)

Abstract

We propose LiDAL, a novel active learning method for 3D LiDAR semantic segmentation by exploiting inter-frame uncertainty among LiDAR frames. Our core idea is that a well-trained model should generate robust results irrespective of viewpoints for scene scanning and thus the inconsistencies in model predictions across frames provide a very reliable measure of uncertainty for active sample selection. To implement this uncertainty measure, we introduce new inter-frame divergence and entropy formulations, which serve as the metrics for active selection. Moreover, we demonstrate additional performance gains by predicting and incorporating pseudo-labels, which are also selected using the proposed inter-frame uncertainty measure. Experimental results validate the effectiveness of LiDAL: we achieve 95% of the performance of fully supervised learning with less than 5% of annotations on the SemanticKITTI and nuScenes datasets, outperforming state-of-the-art active learning methods. Code release: https://github.com/hzykent/LiDAL.

Original language	English
Title of host publication	Computer Vision – ECCV 2022
Subtitle of host publication	17th European Conference, 2022, Proceedings
Editors	Shai Avidan, Gabriel Brostow, Moustapha Cissé, Giovanni Maria Farinella, Tal Hassner
Publisher	Springer, Cham
Pages	248-265
Edition	1
ISBN (Electronic)	978-3-031-19812-0
ISBN (Print)	978-3-031-19811-3
DOIs	https://doi.org/10.1007/978-3-031-19812-0_15
Publication status	Published - 2022
Event	17th European Conference on Computer Vision (ECCV 2022) - Hybrid, Tel-Aviv, Israel Duration: 23 Oct 2022 → 27 Oct 2022 https://eccv2022.ecva.net/

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13687 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th European Conference on Computer Vision (ECCV 2022)
Abbreviated title	ECCV’22
Place	Israel
City	Tel-Aviv
Period	23/10/22 → 27/10/22
Internet address	https://eccv2022.ecva.net/

Bibliographical note

Full text of this publication does not contain sufficient affiliation information. With consent from the author(s) concerned, the Research Unit(s) information for this record is based on the existing academic department affiliation of the author(s).

Funding

This work is supported by Hong Kong RGC GRF 16206722 and a grant from City University of Hong Kong (Project No. 7005729).

Research Keywords

3D LiDAR Semantic Segmentation
Active learning

RGC Funding Information

RGC-funded

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access Output

10.1007/978-3-031-19812-0_15

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Hu, Z., Bai, X., Zhang, R., Wang, X., Sun, G., Fu, H., & Tai, C.-L. (2022). LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation. In S. Avidan, G. Brostow, M. Cissé, G. M. Farinella, & T. Hassner (Eds.), Computer Vision – ECCV 2022: 17th European Conference, 2022, Proceedings (1 ed., pp. 248-265). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13687 LNCS). Springer, Cham. https://doi.org/10.1007/978-3-031-19812-0_15

Hu, Zeyu ; Bai, Xuyang ; Zhang, Runze et al. / LiDAL : Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation. Computer Vision – ECCV 2022: 17th European Conference, 2022, Proceedings. editor / Shai Avidan ; Gabriel Brostow ; Moustapha Cissé ; Giovanni Maria Farinella ; Tal Hassner. 1. ed. Springer, Cham, 2022. pp. 248-265 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{56a7f55f22c648bcb0c82b4be4319675,

title = "LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation",

abstract = "We propose LiDAL, a novel active learning method for 3D LiDAR semantic segmentation by exploiting inter-frame uncertainty among LiDAR frames. Our core idea is that a well-trained model should generate robust results irrespective of viewpoints for scene scanning and thus the inconsistencies in model predictions across frames provide a very reliable measure of uncertainty for active sample selection. To implement this uncertainty measure, we introduce new inter-frame divergence and entropy formulations, which serve as the metrics for active selection. Moreover, we demonstrate additional performance gains by predicting and incorporating pseudo-labels, which are also selected using the proposed inter-frame uncertainty measure. Experimental results validate the effectiveness of LiDAL: we achieve 95% of the performance of fully supervised learning with less than 5% of annotations on the SemanticKITTI and nuScenes datasets, outperforming state-of-the-art active learning methods. Code release: https://github.com/hzykent/LiDAL.",

keywords = "3D LiDAR Semantic Segmentation, Active learning",

author = "Zeyu Hu and Xuyang Bai and Runze Zhang and Xin Wang and Guangyuan Sun and Hongbo Fu and Chiew-Lan Tai",

note = "Full text of this publication does not contain sufficient affiliation information. With consent from the author(s) concerned, the Research Unit(s) information for this record is based on the existing academic department affiliation of the author(s).; 17th European Conference on Computer Vision (ECCV 2022), ECCV{\textquoteright}22 ; Conference date: 23-10-2022 Through 27-10-2022",

year = "2022",

doi = "10.1007/978-3-031-19812-0_15",

language = "English",

isbn = "978-3-031-19811-3",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer, Cham",

pages = "248--265",

editor = "Shai Avidan and Gabriel Brostow and Moustapha Ciss{\'e} and Farinella, {Giovanni Maria} and Tal Hassner",

booktitle = "Computer Vision – ECCV 2022",

edition = "1",

url = "https://eccv2022.ecva.net/",

}

Hu, Z, Bai, X, Zhang, R, Wang, X, Sun, G, Fu, H & Tai, C-L 2022, LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation. in S Avidan, G Brostow, M Cissé, GM Farinella & T Hassner (eds), Computer Vision – ECCV 2022: 17th European Conference, 2022, Proceedings. 1 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13687 LNCS, Springer, Cham, pp. 248-265, 17th European Conference on Computer Vision (ECCV 2022), Tel-Aviv, Israel, 23/10/22. https://doi.org/10.1007/978-3-031-19812-0_15

LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation. / Hu, Zeyu; Bai, Xuyang; Zhang, Runze et al.
Computer Vision – ECCV 2022: 17th European Conference, 2022, Proceedings. ed. / Shai Avidan; Gabriel Brostow; Moustapha Cissé; Giovanni Maria Farinella; Tal Hassner. 1. ed. Springer, Cham, 2022. p. 248-265 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13687 LNCS).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - LiDAL

T2 - 17th European Conference on Computer Vision (ECCV 2022)

AU - Hu, Zeyu

AU - Bai, Xuyang

AU - Zhang, Runze

AU - Wang, Xin

AU - Sun, Guangyuan

AU - Fu, Hongbo

AU - Tai, Chiew-Lan

N1 - Full text of this publication does not contain sufficient affiliation information. With consent from the author(s) concerned, the Research Unit(s) information for this record is based on the existing academic department affiliation of the author(s).

PY - 2022

Y1 - 2022

N2 - We propose LiDAL, a novel active learning method for 3D LiDAR semantic segmentation by exploiting inter-frame uncertainty among LiDAR frames. Our core idea is that a well-trained model should generate robust results irrespective of viewpoints for scene scanning and thus the inconsistencies in model predictions across frames provide a very reliable measure of uncertainty for active sample selection. To implement this uncertainty measure, we introduce new inter-frame divergence and entropy formulations, which serve as the metrics for active selection. Moreover, we demonstrate additional performance gains by predicting and incorporating pseudo-labels, which are also selected using the proposed inter-frame uncertainty measure. Experimental results validate the effectiveness of LiDAL: we achieve 95% of the performance of fully supervised learning with less than 5% of annotations on the SemanticKITTI and nuScenes datasets, outperforming state-of-the-art active learning methods. Code release: https://github.com/hzykent/LiDAL.

AB - We propose LiDAL, a novel active learning method for 3D LiDAR semantic segmentation by exploiting inter-frame uncertainty among LiDAR frames. Our core idea is that a well-trained model should generate robust results irrespective of viewpoints for scene scanning and thus the inconsistencies in model predictions across frames provide a very reliable measure of uncertainty for active sample selection. To implement this uncertainty measure, we introduce new inter-frame divergence and entropy formulations, which serve as the metrics for active selection. Moreover, we demonstrate additional performance gains by predicting and incorporating pseudo-labels, which are also selected using the proposed inter-frame uncertainty measure. Experimental results validate the effectiveness of LiDAL: we achieve 95% of the performance of fully supervised learning with less than 5% of annotations on the SemanticKITTI and nuScenes datasets, outperforming state-of-the-art active learning methods. Code release: https://github.com/hzykent/LiDAL.

KW - 3D LiDAR Semantic Segmentation

KW - Active learning

UR - http://www.scopus.com/inward/record.url?scp=85142676960&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85142676960&origin=recordpage

U2 - 10.1007/978-3-031-19812-0_15

DO - 10.1007/978-3-031-19812-0_15

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 978-3-031-19811-3

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 248

EP - 265

BT - Computer Vision – ECCV 2022

A2 - Avidan, Shai

A2 - Brostow, Gabriel

A2 - Cissé, Moustapha

A2 - Farinella, Giovanni Maria

A2 - Hassner, Tal

PB - Springer, Cham

Y2 - 23 October 2022 through 27 October 2022

ER -

Hu Z, Bai X, Zhang R, Wang X, Sun G, Fu H et al. LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation. In Avidan S, Brostow G, Cissé M, Farinella GM, Hassner T, editors, Computer Vision – ECCV 2022: 17th European Conference, 2022, Proceedings. 1 ed. Springer, Cham. 2022. p. 248-265. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Epub 2022 Oct 30. doi: 10.1007/978-3-031-19812-0_15

LiDAL: Inter-frame Uncertainty Based Active Learning for 3D LiDAR Semantic Segmentation

Abstract

Publication series

Conference

Bibliographical note

Funding

Research Keywords

RGC Funding Information

UN SDGs

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this