Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation

Weakly-supervised point cloud segmentation with extremely limited labels is highly desirable to alleviate the expensive costs of collecting densely annotated 3D points. This paper explores applying the consistency regularization that is commonly used in weakly-supervised learning, for its point clou...

Full description

Saved in:
Bibliographic Details
Main Authors: Wu, Zhonghua, Wu, Yicheng, Lin, Guosheng, Cai, Jianfei
Other Authors: School of Computer Science and Engineering
Format: Article
Language:English
Published: 2024
Subjects:
Online Access:https://hdl.handle.net/10356/176275
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-176275
record_format dspace
spelling sg-ntu-dr.10356-1762752024-05-14T07:33:57Z Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation Wu, Zhonghua Wu, Yicheng Lin, Guosheng Cai, Jianfei School of Computer Science and Engineering Computer and Information Science Weakly supervision Point cloud Weakly-supervised point cloud segmentation with extremely limited labels is highly desirable to alleviate the expensive costs of collecting densely annotated 3D points. This paper explores applying the consistency regularization that is commonly used in weakly-supervised learning, for its point cloud counterpart with multiple data-specific augmentations, which has not been well studied. We observe that the straightforward way of applying consistency constraints to weakly-supervised point cloud segmentation has two major limitations: noisy pseudo labels due to the conventional confidence-based selection and insufficient consistency constraints due to discarding unreliable pseudo labels. Therefore, we propose a novel Reliability-Adaptive Consistency Network (RAC-Net) to use both prediction confidence and model uncertainty to measure the reliability of pseudo labels and apply consistency training on all unlabeled points while with different consistency constraints for different points based on the reliability of corresponding pseudo labels. Experimental results on the S3DIS and ScanNet-v2 benchmark datasets show that our model achieves superior performance in weakly-supervised point cloud segmentation. The code will be released publicly at https://github.com/wu-zhonghua/RAC-Net . Agency for Science, Technology and Research (A*STAR) This research is supported by the Agency for Science, Technology and Research (A*STAR) under its MTC Programmatic Funds (Grant No. M23L7b0021). 2024-05-14T07:33:56Z 2024-05-14T07:33:56Z 2024 Journal Article Wu, Z., Wu, Y., Lin, G. & Cai, J. (2024). Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation. International Journal of Computer Vision. https://dx.doi.org/10.1007/s11263-023-01975-8 0920-5691 https://hdl.handle.net/10356/176275 10.1007/s11263-023-01975-8 2-s2.0-85182237233 en M23L7b0021 International Journal of Computer Vision © 2024 The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature. All rights reserved.
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic Computer and Information Science
Weakly supervision
Point cloud
spellingShingle Computer and Information Science
Weakly supervision
Point cloud
Wu, Zhonghua
Wu, Yicheng
Lin, Guosheng
Cai, Jianfei
Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation
description Weakly-supervised point cloud segmentation with extremely limited labels is highly desirable to alleviate the expensive costs of collecting densely annotated 3D points. This paper explores applying the consistency regularization that is commonly used in weakly-supervised learning, for its point cloud counterpart with multiple data-specific augmentations, which has not been well studied. We observe that the straightforward way of applying consistency constraints to weakly-supervised point cloud segmentation has two major limitations: noisy pseudo labels due to the conventional confidence-based selection and insufficient consistency constraints due to discarding unreliable pseudo labels. Therefore, we propose a novel Reliability-Adaptive Consistency Network (RAC-Net) to use both prediction confidence and model uncertainty to measure the reliability of pseudo labels and apply consistency training on all unlabeled points while with different consistency constraints for different points based on the reliability of corresponding pseudo labels. Experimental results on the S3DIS and ScanNet-v2 benchmark datasets show that our model achieves superior performance in weakly-supervised point cloud segmentation. The code will be released publicly at https://github.com/wu-zhonghua/RAC-Net .
author2 School of Computer Science and Engineering
author_facet School of Computer Science and Engineering
Wu, Zhonghua
Wu, Yicheng
Lin, Guosheng
Cai, Jianfei
format Article
author Wu, Zhonghua
Wu, Yicheng
Lin, Guosheng
Cai, Jianfei
author_sort Wu, Zhonghua
title Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation
title_short Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation
title_full Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation
title_fullStr Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation
title_full_unstemmed Reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation
title_sort reliability-adaptive consistency regularization for weakly-supervised point cloud segmentation
publishDate 2024
url https://hdl.handle.net/10356/176275
_version_ 1800916306537480192