Query-aware Hub Prototype Learning for Few-Shot 3D Point Cloud Semantic Segmentation

YiLin Zhou, Lili Wei, Zheming Xu, Ziyi Chen, Congyan Lang

Keywords: few-shot 3D point cloud semantic segmentation，hubness learning，prototype learning

Abstract: Few-shot 3D point cloud semantic segmentation (FS-3DSeg) aims to segment novel classes with only a few labeled samples. However, existing metric-based prototype learning methods generate prototypes solely from the support set. This often results in prototype bias, where prototypes overfit support-specific characteristics and fail to generalize to the query distribution, especially in the presence of distribution shifts, which leads to degraded segmentation performance. Although some works make efforts to refine or align these prototypes with queries, prototype bias remains poorly addressed, as the initial prototypes have already deviated significantly from the queries. To address this issue, we propose a novel Query-aware Hub Prototype (QHP) learning method that explicitly models semantic correlations between support and query sets. Specifically, we propose a Hub Prototype Generation (HPG) module that constructs a bipartite graph connecting query and support points, identifies frequently linked support hubs, and generates query-relevant prototypes that better capture cross-set semantics. To further mitigate the influence of bad hubs and ambiguous prototypes near class boundaries, we introduce a Prototype Distribution Optimization (PDO) module, which employs a purity-reweighted contrastive loss to refine prototype representations by pulling bad hubs and outlier prototypes closer to their corresponding class centers. Extensive experiments on S3DIS and ScanNet demonstrate that QHP achieves substantial performance gains over state-of-the-art methods, effectively narrowing the semantic gap between prototypes and query sets in FS-3DSeg.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 8774