Adaptive Margin Contrastive Learning for Ambiguity-aware 3D Semantic Segmentation

TL;DR

This paper introduces AMContrast3D, an adaptive margin contrastive learning approach that accounts for point ambiguity in 3D semantic segmentation, improving accuracy by customizing constraints based on ambiguity levels.

Contribution

The paper presents a novel adaptive margin contrastive learning framework that dynamically adjusts decision boundaries according to point ambiguity, addressing limitations of fixed-margin methods.

Findings

Outperforms state-of-the-art on S3DIS and ScanNet datasets.

Effectively handles ambiguous points with adaptive margins.

Improves segmentation accuracy in transition regions.

Abstract

In this paper, we propose an adaptive margin contrastive learning method for 3D point cloud semantic segmentation, namely AMContrast3D. Most existing methods use equally penalized objectives, which ignore per-point ambiguities and less discriminated features stemming from transition regions. However, as highly ambiguous points may be indistinguishable even for humans, their manually annotated labels are less reliable, and hard constraints over these points would lead to sub-optimal models. To address this, we design adaptive objectives for individual points based on their ambiguity levels, aiming to ensure the correctness of low-ambiguity points while allowing mistakes for high-ambiguity points. Specifically, we first estimate ambiguities based on position embeddings. Then, we develop a margin generator to shift decision boundaries for contrastive feature embeddings, so margins are narrowed due to increasing ambiguities with even negative margins for extremely high-ambiguity points. Experimental results on large-scale datasets, S3DIS and ScanNet, demonstrate that our method outperforms state-of-the-art methods.