Generalized Few-Shot Point Cloud Segmentation Via Geometric Words

TL;DR

This paper introduces a generalized few-shot point cloud segmentation method using geometric words and prototypes, enabling better adaptation to new classes while retaining base class segmentation, demonstrated on S3DIS and ScanNet datasets.

Contribution

It proposes a novel geometric-aware semantic representation and geometric prototypes to improve few-shot segmentation of point clouds, addressing the challenge of class generalization and retention.

Findings

Superior performance over baselines on S3DIS and ScanNet datasets

Effective generalization to new classes with limited support data

Retains segmentation accuracy for base classes

Abstract

Existing fully-supervised point cloud segmentation methods suffer in the dynamic testing environment with emerging new classes. Few-shot point cloud segmentation algorithms address this problem by learning to adapt to new classes at the sacrifice of segmentation accuracy for the base classes, which severely impedes its practicality. This largely motivates us to present the first attempt at a more practical paradigm of generalized few-shot point cloud segmentation, which requires the model to generalize to new categories with only a few support point clouds and simultaneously retain the capability to segment base classes. We propose the geometric words to represent geometric components shared between the base and novel classes, and incorporate them into a novel geometric-aware semantic representation to facilitate better generalization to the new classes without forgetting the old ones. Moreover, we introduce geometric prototypes to guide the segmentation with geometric prior knowledge. Extensive experiments on S3DIS and ScanNet consistently illustrate the superior performance of our method over baseline methods. Our code is available at: https://github.com/Pixie8888/GFS-3DSeg_GWs.