PointDAN: A Multi-Scale 3D Domain Adaption Network for Point Cloud Representation

Abstract

Domain Adaptation (DA) approaches achieved significant improvements in a wide range of machine learning and computer vision tasks (i.e., classification, detection, and segmentation). However, as far as we are aware, there are few methods yet to achieve domain adaptation directly on 3D point cloud data. The unique challenge of point cloud data lies in its abundant spatial geometric information, and the semantics of the whole object is contributed by including regional geometric structures. Specifically, most general-purpose DA methods that struggle for global feature alignment and ignore local geometric information are not suitable for 3D domain alignment. In this paper, we propose a novel 3D Domain Adaptation Network for point cloud data (PointDAN). PointDAN jointly aligns the global and local features in multi-level. For local alignment, we propose Self-Adaptive (SA) node module with an adjusted receptive field to model the discriminative local structures for aligning domains. To represent hierarchically scaled features, node-attention module is further introduced to weight the relationship of SA nodes across objects and domains. For global alignment, an adversarial-training strategy is employed to learn and align global features across domains. Since there is no common evaluation benchmark for 3D point cloud DA scenario, we build a general benchmark (i.e., PointDA-10) extracted from three popular 3D object/scene datasets (i.e., ModelNet, ShapeNet and ScanNet) for cross-domain 3D objects classification fashion. Extensive experiments on PointDA-10 illustrate the superiority of our model over the state-of-the-art general-purpose DA methods.

12 pages, 4 figures, 33rd Conference on Neural Information Processing Systems (NeurIPS 2019)

Cited by in corpus (26)

• Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation
• A Survey on Deep Domain Adaptation for LiDAR Perception
• Deep Learning for 3D Point Clouds: A Survey
• Self-Supervised Learning for Domain Adaptation on Point-Clouds
• Deformable PV-RCNN: Improving 3D Object Detection with Learned Deformations
• What Can Be Transferred: Unsupervised Domain Adaptation for Endoscopic Lesions Segmentation
• Complete & Label: A Domain Adaptation Approach to Semantic Segmentation of LiDAR Point Clouds
• See Eye to Eye: A Lidar-Agnostic 3D Detection Framework for Unsupervised Multi-Target Domain Adaptation
• Uncertainty-aware Mean Teacher for Source-free Unsupervised Domain Adaptive 3D Object Detection
• Adapting Object Detectors with Conditional Domain Normalization
• Unsupervised Point Cloud Pre-Training via Occlusion Completion
• LiDARNet: A Boundary-Aware Domain Adaptation Model for Point Cloud Semantic Segmentation
• A Learnable Self-supervised Task for Unsupervised Domain Adaptation on Point Clouds
• SPG: Unsupervised Domain Adaptation for 3D Object Detection via Semantic Point Generation
• SF-UDA${}^{3D}$: Source-Free Unsupervised Domain Adaptation for LiDAR-Based 3D Object Detection
• VDM-DA: Virtual Domain Modeling for Source Data-free Domain Adaptation
• ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection
• Exploiting Playbacks in Unsupervised Domain Adaptation for 3D Object Detection
• Attention Models for Point Clouds in Deep Learning: A Survey
• Contradictory Structure Learning for Semi-supervised Domain Adaptation
• Generation For Adaption: A GAN-Based Approach for 3D Domain Adaption with Point Cloud Data
• Point Discriminative Learning for Data-efficient 3D Point Cloud Analysis
• Look, Cast and Mold: Learning 3D Shape Manifold from Single-view Synthetic Data
• R-AGNO-RPN: A LIDAR-Camera Region Deep Network for Resolution-Agnostic Detection
• Geometry-Aware Self-Training for Unsupervised Domain Adaptationon Object Point Clouds
• Classifying In-Place Gestures with End-to-End Point Cloud Learning