Research map / Detection and segmentation / 3D detection and driving
Motal: Unsupervised 3D Object Detection by Modality and Task-specific Knowledge Transfer
Hai Wu, Hongwei Lin, Xusheng Guo, Xin Li, Mingming Wang, Cheng Wang, Chenglu Wen
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 Modality-specific Label Extension (MLE) (ablation) | Motal (CPD) · WOD val | mAP L2 AP: 16.74 -> 29.74 | Improved |
| add Task-specific Masked Training (TMT) on top of MLE (ablation) | Motal (CPD) · WOD val | mAP L2 AP: 29.74 -> 34.34 | Improved |
| add TMT without MLE (ablation) | Motal (CPD) · WOD val | mAP L2 AP: 16.74 -> 25.02 | Improved |
| add class generalization loss (ablation) | MLE · WOD val | mAP L2 AP: 28.41 -> 34.34 | Improved |
| add box propagation (ablation) | MLE · WOD val | mAP L2 AP: 24.34 -> 34.34 | Improved |
| add soft IoU loss (ablation) | MLE · WOD val | mAP L2 AP: 33.21 -> 34.34 | Improved |
| add soft IoU scoring (ablation) | MLE · WOD val | mAP L2 AP: 32.82 -> 34.34 | Improved |
| use TMT instead of high classification score threshold (ablation) | Motal (CPD) · WOD val | mAP L2 AP: 24.33 -> 34.34 | Improved |
| use TMT instead of high thresholds on both scores (ablation) | Motal (CPD) · WOD val | mAP L2 AP: 28.89 -> 34.34 | Improved |
| swap ViT-S heatmap backbone for Swin-S (ablation) | Motal (CPD) · WOD val | mAP L2 AP: 34.34 -> 33.71 | No clear effect |
| swap ViT-S heatmap backbone for ResNet (ablation) | Motal (CPD) · WOD val | mAP L2 AP: 34.34 -> 34.15 | No clear effect |
| train directly on CPD pseudo labels | CPD pseudo labels · WOD val | 3D AP L2 pedestrian: 13.22 -> 7.09 | Hurt |
Closest papers by meaning
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.87)
- SP3D: Boosting Sparsely-Supervised 3D Object Detection via Accurate Cross-Modal Semantic Prompts (CVPR 2025 · similarity 0.87)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · similarity 0.87)
- Dual-Perspective Knowledge Enrichment for Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.86)
- UniDet3D: Multi-dataset Indoor 3D Object Detection (AAAI 2025 · similarity 0.86)
- LabelAny3D: Label Any Object 3D in the Wild (NeurIPS 2025 · similarity 0.86)
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.
