Unsupervised Domain Adaptation for 3D LiDAR Semantic Segmentation Using Contrastive Learning and Multi-Model Pseudo Labeling

TL;DR

This paper proposes a novel unsupervised domain adaptation framework for 3D LiDAR semantic segmentation that combines contrastive learning and multi-model pseudo-labeling to improve performance across different domains without target labels.

Contribution

It introduces a two-stage approach using contrastive pre-training and ensemble pseudo-labeling with multiple architectures for effective domain adaptation.

Findings

Significant accuracy improvements on SemanticKITTI to SemanticPOSS and SemanticSlamantic datasets.

Effective reduction of domain gap without target domain annotations.

Enhanced robustness through multi-model pseudo-label aggregation.

Abstract

Addressing performance degradation in 3D LiDAR semantic segmentation due to domain shifts (e.g., sensor type, geographical location) is crucial for autonomous systems, yet manual annotation of target data is prohibitive. This study addresses the challenge using Unsupervised Domain Adaptation (UDA) and introduces a novel two-stage framework to tackle it. Initially, unsupervised contrastive learning at the segment level is used to pre-train a backbone network, enabling it to learn robust, domain-invariant features without labels. Subsequently, a multi-model pseudo-labeling strategy is introduced, utilizing an ensemble of diverse state-of-the-art architectures (including projection, voxel, hybrid, and cylinder-based methods). Predictions from these models are aggregated via hard voting to generate high-quality, refined pseudo-labels for the unlabeled target domain, mitigating single-model biases. The contrastively pre-trained network is then fine-tuned using these robust pseudo-labels. Experiments adapting from SemanticKITTI to unlabeled target datasets (SemanticPOSS, SemanticSlamantic) demonstrate significant improvements in segmentation accuracy compared to direct transfer and single-model UDA approaches. These results highlight the effectiveness of combining contrastive pre-training with refined ensemble pseudo-labeling for bridging complex domain gaps without requiring target domain annotations.