Research map / Detection and segmentation / 3D detection and driving
SPG: Unsupervised Domain Adaptation for 3D Object Detection via Semantic Point Generation
Qiangeng Xu, Yin Zhou, Weiyue Wang, Charles R. Qi, Dragomir Anguelov
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Hide and Predict training (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 39.22 -> 41.56 | Improved |
| add Semantic Area Expansion (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 38.09 -> 41.56 | Improved |
| add foreground confidence as point property (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 37.96 -> 41.56 | Improved |
| use expansion weight alpha=0.5 instead of 1.0 (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 38.42 -> 41.56 | Improved |
| use expansion weight alpha=0.5 instead of 0.0 (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 38.96 -> 41.56 | Improved |
| use P_thresh 0.5 instead of 0.3 (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 39.39 -> 41.56 | Improved |
| use P_thresh 0.5 instead of 0.7 (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 40.89 -> 41.56 | Improved |
| add foreground confidence only (no hide, no expansion) (ablation) | SPG + PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 34.65 -> 35.89 | Improved |
| add SPG semantic points | PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 34.65 -> 41.56 | Improved |
| randomly drop 17% of source points (RndDrop) | PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 34.65 -> 35.45 | Improved |
| aggregate 5 frames | PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 34.65 -> 38.51 | Improved |
| add adversarial domain classification loss (ADA) | PointPillars · Waymo Kirk val, vehicle LEVEL_1 (trained on OD) | 3D AP: 34.65 -> 36.34; source AP drop of 1.52 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Voxel R-CNN — Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection (AAAI 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SSDA3D: Semi-supervised Domain Adaptation for 3D Object Detection from Point Cloud (AAAI 2023 · similarity 0.91)
- Towards Universal LiDAR-Based 3D Object Detection by Multi-Domain Knowledge Transfer (ICCV 2023 · similarity 0.90)
- Towards Unsupervised Object Detection From LiDAR Point Clouds (CVPR 2023 · similarity 0.90)
- GBlobs: Explicit Local Structure via Gaussian Blobs for Improved Cross-Domain LiDAR-based 3D Object Detection (CVPR 2025 · similarity 0.89)
- CMDA: Cross-Modal and Domain Adversarial Adaptation for LiDAR-Based 3D Object Detection (AAAI 2024 · similarity 0.89)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
