Research map / Detection and segmentation / 3D detection and driving
Unsupervised Domain Adaptation for Monocular 3D Object Detection via Self-Training
Zhenyu Li, Zehui Chen, Ang Li, Liangji Fang, Qinhong Jiang, Xianming Liu, Junjun Jiang
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 geometry-aligned multi-scale training (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 0 -> 21.89 | Improved |
| add GAMS to naive self-training (ablation) | Naive ST · NuScenes->KITTI | AP40 3D Moderate: 0 -> 3.69 | Improved |
| use end-to-end mean teacher instead of naive ST (both with GAMS) (ablation) | STMono3D · NuScenes->KITTI | AP40 3D Moderate: 3.69 -> 16.85 | Improved |
| add EMA teacher (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 0.82 -> 21.89 | Improved |
| add QAS reweighting on classification loss (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 18.14 -> 21.89 | Improved |
| add QAS reweighting on regression loss (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 18.14 -> 13.80 | Hurt |
| apply QAS to both cls and reg losses (vs cls only) (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 21.89 -> 15.67 | Hurt |
| add positive focusing training (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 7.80 -> 13.15 | Improved |
| add dynamic threshold (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 7.80 -> 13.73 | Improved |
| combine PFT and DT (ablation) | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 7.80 -> 21.89 | Improved |
| train with STMono3D vs oracle source-trained model | STMono3D · NuScenes->KITTI | AP11 3D Moderate: 19.88 -> 21.89 | Improved |
| train with STMono3D vs oracle | STMono3D · NuScenes->KITTI | AP40 3D Moderate: 18.97 -> 16.85 | Hurt |
Closest papers by meaning
- ST3D: Self-Training for Unsupervised Domain Adaptation on 3D Object Detection (CVPR 2021 · similarity 0.93)
- Geometry-Guided Domain Generalization for Monocular 3D Object Detection (AAAI 2024 · similarity 0.91)
- Unified Domain Generalization and Adaptation for Multi-View 3D Object Detection (NeurIPS 2024 · similarity 0.90)
- Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D Object Detection (CVPR 2024 · similarity 0.90)
- MonoTAKD: Teaching Assistant Knowledge Distillation for Monocular 3D Object Detection (CVPR 2025 · similarity 0.90)
- MonoDiff: Monocular 3D Object Detection and Pose Estimation with Diffusion Models (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.
