Cross-modal knowledge transfer for 3D point clouds via graph offset prediction

Open Webpage

Huang Zhang, Long Yu, Guoqi Wang, Shengwei Tian, Zaiyang Yu, Weijun Li, Xin Ning

Published: 01 Jan 2025, Last Modified: 09 Nov 2025Pattern Recognit. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•A graph offset-based encoder, GoConv, is proposed to enhance the local structure capture of point clouds.•Introducing a joint point-view pretraining strategy to learn a unified representation of two modalities.•Assign point cloud and view information to the same feature space through cross-modal alignment.•Validation of the proposed method in standard and zero-shot classification tasks.•Excellent performance of the proposed method in 3D classification tasks without textual information.