A learnable self-supervised task for unsupervised domain adaptation on point cloud classification and segmentation

作     者：Shaolei LIU Xiaoyuan LUO Kexue FU Manning WANG Zhijian SONG Shaolei LIU;Xiaoyuan LUO;Kexue FU;Manning WANG;Zhijian SONG

作者机构：Shanghai Key Laboratory of Medical Image Computing and Computer Assisted InterventionShanghai 200032China Digital Medical Research CenterSchool of Basic Medical ScienceFudan UniversityShanghai 200032China 

出 版 物：《Frontiers of Computer Science》 (中国计算机科学前沿（英文版）)

年 卷 期：2023年第17卷第6期

页      面：147-149页

核心收录：

学科分类：08[工学] 081104[工学-模式识别与智能系统] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported by the National Natural Science Foundation of China(Grant No.62076070). 

主　　题：point cloud utilizing 

摘      要：1 Introduction Deep neural networks have exhibited excellent performance in supervised tasks on point clouds,such as classification,segmentation[1]and registration[2].In supervised learning schemes,manual labeling of massive point clouds is needed for model training.However,point clouds captured from different scenarios exist inevitable distribution discrepancy,and model trained from one domain always generalize badly in another domain.To reduce the doamin distribution discrepancy,many studies[3–6]have emerged for point cloud unsupervised domain adaptation(UDA)by learning domain-invariant features,where[5]proposed using adaptive nodes to align the local features between the source and the target domains[3,4],and[6]proposed utilizing self-supervised tasks to help capture highly transferable feature representations.