Research map / Detection and segmentation / 3D detection and driving
Bi3D: Bi-Domain Active Learning for Cross-Domain 3D Object Detection
Jiakang Yuan, Bo Zhang, Xiangchao Yan, Tao Chen, Botian Shi, Yikang Li, Yu Qiao
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 |
|---|---|---|---|
| Use Bi3D with 1% target annotation budget instead of ST3D with SN | PV-RCNN · nuScenes→KITTI | AP_BEV: 84.29 -> 87.00 | Improved |
| Use Bi3D with 1% target annotation budget instead of ST3D with SN | PV-RCNN · Waymo→KITTI | AP_3D: 76.86 -> 78.03 | Improved |
| Increase target annotation budget from 1% to 5% | Bi3D with SN · nuScenes→KITTI | AP_BEV: 87.00 -> 89.63 | Improved |
| Add domainness-aware active source sampling instead of random source sampling (ablation) | Bi3D with SN and random target sampling · nuScenes→KITTI | AP_BEV: 84.42 -> 85.02 | Improved |
| Add diversity-based active target sampling instead of random target sampling (ablation) | Bi3D with SN and random source sampling · nuScenes→KITTI | AP_BEV: 84.42 -> 86.53 | Improved |
| Use active sampling for both source and target rather than random sampling for both (ablation) | Bi3D with SN · Waymo→KITTI | AP_BEV: 85.48 -> 87.12 | Improved |
| Remove SN (ablation) | Bi3D · nuScenes→KITTI | AP_3D: 77.55 -> 71.56 | Hurt |
| Combine objectness and entropy scores for foreground-aware discriminator features (ablation) | Bi3D source-and-target sampling · Waymo→KITTI | AP_3D: 76.69 -> 78.03 | Improved |
| Increase selected source proportion from 1% to 25% (ablation) | Bi3D · Waymo→nuScenes | AP_BEV: 86.95 -> 85.24 | Hurt |
| Combine Bi3D active annotation with ST3D pseudo-labeling of remaining target data (ablation) | PV-RCNN · nuScenes→KITTI | AP_BEV: 87.00 -> 89.28 | Improved |
Closest papers by meaning
- Bi3D++: Hybrid Bi-Domain Active Learning for Cross-Domain 3D Object Detection (TPAMI 2026 · similarity 0.99)
- Unsupervised Domain Adaptation for Monocular 3D Object Detection via Self-Training (ECCV 2022 · similarity 0.89)
- ST3D: Self-Training for Unsupervised Domain Adaptation on 3D Object Detection (CVPR 2021 · similarity 0.89)
- Unified Domain Generalization and Adaptation for Multi-View 3D Object Detection (NeurIPS 2024 · similarity 0.89)
- Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D Object Detection (CVPR 2024 · similarity 0.88)
- 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.
