Abstract:LiDAR-based 3D object detection has seen impressive advances in recent times. However, deploying trained 3D detectors in the real world often yields unsatisfactory performance when the distribution of the test data significantly deviates from the training data due to different weather conditions, object sizes, \textit{etc}. A key factor in this performance degradation is the diminished generalizability of pre-trained models, which creates a sharp loss landscape during training. Such sharpness, when encountered during testing, can precipitate significant performance declines, even with minor data variations. To address the aforementioned challenges, we propose \textbf{dual-perturbation optimization (DPO)} for \textbf{\underline{T}est-\underline{t}ime \underline{A}daptation in \underline{3}D \underline{O}bject \underline{D}etection (TTA-3OD)}. We minimize the sharpness to cultivate a flat loss landscape to ensure model resiliency to minor data variations, thereby enhancing the generalization of the adaptation process. To fully capture the inherent variability of the test point clouds, we further introduce adversarial perturbation to the input BEV features to better simulate the noisy test environment. As the dual perturbation strategy relies on trustworthy supervision signals, we utilize a reliable Hungarian matcher to filter out pseudo-labels sensitive to perturbations. Additionally, we introduce early Hungarian cutoff to avoid error accumulation from incorrect pseudo-labels by halting the adaptation process. Extensive experiments across three types of transfer tasks demonstrate that the proposed DPO significantly surpasses previous state-of-the-art approaches, specifically on Waymo $\rightarrow$ KITTI, outperforming the most competitive baseline by 57.72\% in $\text{AP}_\text{3D}$ and reaching 91\% of the fully supervised upper bound.

GeoAdapt: Self-Supervised Test-Time Adaptation in LiDAR Place Recognition Using Geometric Priors

Domain Adaptation on Point Clouds Via Geometry-Aware Implicits

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

MOS: Model Synergy for Test-Time Adaptation on LiDAR-Based 3D Object Detection

GeoNet: Benchmarking Unsupervised Adaptation across Geographies

GPA-3D: Geometry-aware Prototype Alignment for Unsupervised Domain Adaptive 3D Object Detection from Point Clouds

DASGIL: Domain Adaptation for Semantic and Geometric-Aware Image-Based Localization.

Range Adaptation for 3D Object Detection in LiDAR

LiDAR-UDA: Self-ensembling Through Time for Unsupervised LiDAR Domain Adaptation

Towards Minimizing the LiDAR Sim-to-Real Domain Shift: Object-Level Local Domain Adaptation for 3D Point Clouds of Autonomous Vehicles

Learning-on-the-Drive: Self-supervised Adaptation of Visual Offroad Traversability Models

Unsupervised Domain Adaptation for Self-Driving from Past Traversal Features

Cycle and Semantic Consistent Adversarial Domain Adaptation for Reducing Simulation-to-Real Domain Shift in LiDAR Bird's Eye View

UADA3D: Unsupervised Adversarial Domain Adaptation for 3D Object Detection with Sparse LiDAR and Large Domain Gaps

Geometry-Guided Adaptation for Road Segmentation

HGL: Hierarchical Geometry Learning for Test-time Adaptation in 3D Point Cloud Segmentation

See Eye to Eye: A Lidar-Agnostic 3D Detection Framework for Unsupervised Multi-Target Domain Adaptation

Domain Adaptation in 3D Object Detection with Gradual Batch Alternation Training

Pose-Graph Attentional Graph Neural Network for Lidar Place Recognition

DPO: Dual-Perturbation Optimization for Test-time Adaptation in 3D Object Detection

LiOn-XA: Unsupervised Domain Adaptation via LiDAR-Only Cross-Modal Adversarial Training