Exploring multi-scale and cross-type features in 3D point cloud learning with CCMNET

Open Webpage

Wei Zhou, Weiwei Jin, Dekui Wang, Xingxing Hao, Yongxiang Yu, Caiwen Ma

Published: 2025, Last Modified: 28 Jan 2026Expert Syst. Appl. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Proposes 3D point cloud learning with coarse-to-fine & cross-type multi-scale features.•Extracts mid-grained features to balance voxel/point-based coarse/fine extremes.•Voxel and sphere scales set to capture multi-scale local features across types.•Validated on ModelNet40, ShapeNet Part, S3DIS for classification & segmentation.