MD-Mamba: Feature extractor on 3D representation with multi-view depth

Open Webpage

Qihui Li, Zongtan Li, Lianfang Tian, Qiliang Du, Guoyu Lu

Published: 01 Jan 2025, Last Modified: 12 Apr 2025Image Vis. Comput. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•MD-Mamba improves 3D point cloud feature extraction by integrating multi-view depth maps.•Multi-view depth maps are treated as video streams to capture dynamic features effectively.•Siamese Cluster module optimizes feature clustering, enhancing classification accuracy.•Extensive tests on ModelNet40, ShapeNetPart, and ScanObjectNN confirm MD-Mamba’s effectiveness.