Test-time adaptation for geospatial point cloud semantic segmentation with distinct domain shifts

TL;DR

This paper introduces a test-time adaptation method for geospatial point cloud semantic segmentation that improves model accuracy across various domain shifts without retraining, using batch normalization updates and self-supervised learning.

Contribution

It proposes a novel TTA approach that progressively updates BN statistics and employs self-supervised learning for domain adaptation in geospatial PCSS, addressing multiple domain shift scenarios.

Findings

Improves classification accuracy by up to 20% mIoU.

Achieves 59.46% mIoU and 85.97% OA on photogrammetric to airborne adaptation.

Outperforms existing methods in domain shift scenarios.

Abstract

Domain adaptation (DA) techniques help deep learning models generalize across data shifts for point cloud semantic segmentation (PCSS). Test-time adaptation (TTA) allows direct adaptation of a pre-trained model to unlabeled data during inference stage without access to source data or additional training, avoiding privacy issues and large computational resources. We address TTA for geospatial PCSS by introducing three domain shift paradigms: photogrammetric to airborne LiDAR, airborne to mobile LiDAR, and synthetic to mobile laser scanning. We propose a TTA method that progressively updates batch normalization (BN) statistics with each testing batch. Additionally, a self-supervised learning module optimizes learnable BN affine parameters. Information maximization and reliability-constrained pseudo-labeling improve prediction confidence and supply supervisory signals. Experimental results show our method improves classification accuracy by up to 20\% mIoU, outperforming other methods. For photogrammetric (SensatUrban) to airborne (Hessigheim 3D) adaptation at the inference stage, our method achieves 59.46\% mIoU and 85.97\% OA without retraining or fine-turning.