Research map / Detection and segmentation / Open-set and few-shot detection
Bi3D++: Hybrid Bi-Domain Active Learning for Cross-Domain 3D Object Detection
Jiakang Yuan, Xiangchao Yan, Botian Shi, Bo Zhang, Feng Xu, Yu Qiao, Tao Chen
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 domainness-aware active source sampling (instead of random), with SN (ablation) | Bi3D · nuScenes->KITTI | AP_3D: 58.80 -> 65.40 | Improved |
| add domainness-aware active source sampling (instead of random), with SN (ablation) | Bi3D · Waymo->KITTI | AP_3D: 56.20 -> 67.41 | Improved |
| add diversity-based active target sampling instead of random target sampling (ablation) | Bi3D · nuScenes->KITTI | AP_3D: 75.43 -> 77.55 | Improved |
| add random source sampling without SN (ablation) | PV-RCNN · Waymo->KITTI | AP_3D: 22.01 -> 8.78 | Hurt |
| add statistic normalization (SN) to Bi3D (ablation) | Bi3D · nuScenes->KITTI | AP_3D: 71.56 -> 77.55 | Improved |
| add statistic normalization (SN) to Bi3D (ablation) | Bi3D · Waymo->KITTI | AP_3D: 71.36 -> 78.03 | Improved |
| add objectness and entropy scores to discriminator input (source-only sampling) (ablation) | Bi3D · nuScenes->KITTI | AP_3D: 58.74 -> 65.40 | Improved |
| add objectness and entropy scores to discriminator input (S+T) (ablation) | Bi3D · Waymo->KITTI | AP_3D: 76.69 -> 78.03 | Improved |
| increase annotation budget from 1% to 5% (with SN) | Bi3D · Waymo->KITTI | AP_3D: 78.03 -> 81.32 | Improved |
| sample both source and target domains instead of UDA self-training only (source-domain retention) | Bi3D vs ST3D · nuScenes->KITTI | source-domain AP_3D: 19.88 -> 22.89 | Improved |
| combine Bi3D with ST3D pseudo-labeling | Bi3D · nuScenes->KITTI | AP_3D: 77.55 -> 79.69 | Improved |
| apply 2D ADA methods CLUE/TQS instead of Bi3D (w/ SN, 1%) | TQS · Waymo->Lyft | AP_3D: 58.53 -> 55.25 | Hurt |
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.
- ST3D — ST3D: Self-Training for Unsupervised Domain Adaptation on 3D Object Detection (CVPR 2021) · better on 6 of 6 shared results
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
- Bi3D: Bi-Domain Active Learning for Cross-Domain 3D Object Detection (CVPR 2023 · similarity 0.99)
- ST3D: Self-Training for Unsupervised Domain Adaptation on 3D Object Detection (CVPR 2021 · similarity 0.90)
- Unsupervised Domain Adaptation for Monocular 3D Object Detection via Self-Training (ECCV 2022 · similarity 0.90)
- Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D Object Detection (CVPR 2024 · similarity 0.89)
- Unified Domain Generalization and Adaptation for Multi-View 3D Object Detection (NeurIPS 2024 · similarity 0.89)
- CMDA: Cross-Modal and Domain Adversarial Adaptation for LiDAR-Based 3D Object Detection (AAAI 2024 · similarity 0.87)
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.
