S2AM3D: Scale-controllable Part Segmentation of 3D Point Cloud
This work solves part segmentation for 3D computer vision, offering a novel method with a new dataset, but it is incremental in improving existing approaches.
The paper tackles the problem of part-level point cloud segmentation by addressing generalization and consistency issues, proposing S2AM3D which achieves leading performance with exceptional robustness and controllability in experiments.
Part-level point cloud segmentation has recently attracted significant attention in 3D computer vision. Nevertheless, existing research is constrained by two major challenges: native 3D models lack generalization due to data scarcity, while introducing 2D pre-trained knowledge often leads to inconsistent segmentation results across different views. To address these challenges, we propose S2AM3D, which incorporates 2D segmentation priors with 3D consistent supervision. We design a point-consistent part encoder that aggregates multi-view 2D features through native 3D contrastive learning, producing globally consistent point features. A scale-aware prompt decoder is then proposed to enable real-time adjustment of segmentation granularity via continuous scale signals. Simultaneously, we introduce a large-scale, high-quality part-level point cloud dataset with more than 100k samples, providing ample supervision signals for model training. Extensive experiments demonstrate that S2AM3D achieves leading performance across multiple evaluation settings, exhibiting exceptional robustness and controllability when handling complex structures and parts with significant size variations.