Graph Attentive Dual Ensemble learning for Unsupervised Domain Adaptation on point clouds

Li, Qing1,2; Yan, Chuan3; Hao, Qi2; Peng, Xiaojiang1; Liu, Li4

2024-11-01

10.1016/j.patcog.2024.110690

PATTERN RECOGNITION   影响因子和分区

0031-3203

1873-5142

Due to the annotation difficulty of point clouds, Unsupervised Domain Adaptation (UDA) is a promising direction to address unlabeled point cloud classification and segmentation. Recent works show that adding a self -supervised learning branch for target domain training consistently boosts UDA point cloud tasks. However, most of these works simply resort to geometric deformation, which ignores semantic information and is hard to bridge the domain gap. In this paper, we propose a novel self -learning strategy for UDA on point clouds, termed as Graph Attentive Dual Ensemble learning (GRADE), which delivers semantic information directly. Specifically, with a pre -training process on the source domain, GRADE further builds dual collaborative training branches on the target domain, where each of them constructs a temporal average teacher model and distills its pseudo labels to the other branch. To achieve faithful labels from each teacher model, we improve the popular DGCNN architecture by introducing a dynamic graph attentive module to mine the relation between local neighborhood points. We conduct extensive experiments on several UDA point cloud benchmarks, and the results demonstrate that our GRADE method outperforms the state-of-the-art methods on both classification and segmentation tasks with clear margins.

Point clouds Unsupervised Domain Adaptation Dual ensemble learning Graph attentive module Pseudo labels

SCI ; EI

英语

其他

Shenzhen Technology University School-Level Research Project[20231063010070] ; National Natural Science Foundation of China[62261160654] ; Stable Support Projects for Shenzhen Higher Education Institutions[20220718110 918001] ; Natural Science Foundation of Top Talent of SZTU[GD RC202131]

Computer Science ; Engineering

Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic

WOS:001260836300001

ELSEVIER SCI LTD

20242616423272

Personnel training ; Semantics

Personnel:912.4

ENGINEERING

Web of Science

1.Shenzhen Technol Univ, Coll Big data & Internet, Shenzhen 518118, Peoples R China
2.Southern Univ Sci & Technol, Sch Comp Sci & Engn, Shenzhen 518055, Guangdong, Peoples R China
3.George Mason Univ, Comp Sci & Engn, Fairfax, VA USA
4.Natl Univ Def Technol, Coll Elect Sci & Technol, Changsha, Peoples R China

Li, Qing,Yan, Chuan,Hao, Qi,et al. Graph Attentive Dual Ensemble learning for Unsupervised Domain Adaptation on point clouds[J]. PATTERN RECOGNITION,2024,155.

Li, Qing,Yan, Chuan,Hao, Qi,Peng, Xiaojiang,&Liu, Li.(2024).Graph Attentive Dual Ensemble learning for Unsupervised Domain Adaptation on point clouds.PATTERN RECOGNITION,155.

Li, Qing,et al."Graph Attentive Dual Ensemble learning for Unsupervised Domain Adaptation on point clouds".PATTERN RECOGNITION 155(2024).