Uncertainty-aware map-level point cloud semantic segmentation

Weijian Zhang; Zhenping Sun; Hao Fu

doi:10.1117/12.3040309

4 September 2024 Uncertainty-aware map-level point cloud semantic segmentation

Weijian Zhang, Zhenping Sun, Hao Fu

Proceedings Volume 13259, International Conference on Automation Control, Algorithm, and Intelligent Bionics (ACAIB 2024); 132592B (2024) https://doi.org/10.1117/12.3040309
Event: Fourth International Conference on Automation Control, Algorithm, and Intelligent Bionics (ICAIB 2024), 2024, Yinchuan, China

Abstract

In this paper, we propose an uncertainty-aware map-level point cloud semantic segmentation method. We divide the point cloud map into multiple small blocks by a specific strategy, which serve as the input of an uncertainty-aware backbone. The backbone can predicts both the semantic segmentation label and the uncertainty of the prediction. During training, the uncertainty-aware backbone first extracts the high-level features of point cloud through an encoder. Then the high-level features are fed into the learnable normalizing-flow module to approximate the dirichlet prior distribution. Both the parameters of the encoder and the normalizing-flow module can be updated during gradient backpropogation. During inference, the backbone predicts the dirichlet distribution parameter, which can be used for joint computation of semantic segmentation label and the uncertainty of the prediction. Experimental results show that our method has good performance in semantic segmentation and uncertainty prediction for unseen samples.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Weijian Zhang, Zhenping Sun, and Hao Fu "Uncertainty-aware map-level point cloud semantic segmentation", Proc. SPIE 13259, International Conference on Automation Control, Algorithm, and Intelligent Bionics (ACAIB 2024), 132592B (4 September 2024); https://doi.org/10.1117/12.3040309

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
6 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Semantics

Point clouds

Data modeling

Feature extraction

Education and training

LIDAR

Neural networks

Show All Keywords

Keywords/Phrases

Search In:

Publication Years