Research map / Detection and segmentation / 3D detection and driving
CLIPDet3D: Vision-Language Collaborative Distillation for 3D Object Detection
Jiaqi Zhao, Huanfeng Hu, Yong Zhou, Wen-Liang Du, Kunyang Sun, Rui Yao, Qigong Sun
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 Vision-Language Collaborative Learning (ablation) | GeoBEV · nuScenes val | mAP: 0.415 -> 0.426 | Improved |
| add Depth Feature Contrastive Distillation (ablation) | GeoBEV + VLCL · nuScenes val | mAP: 0.426 -> 0.430 | Improved |
| add Dual-Stream Prompt Attention (ablation) | GeoBEV + VLCL + DFCD · nuScenes val | mAP: 0.430 -> 0.433 | Improved |
| add semantic loss Lsem to VLCL (ablation) | BEVDet + VLCL · nuScenes val | rc-mAP: 0.151 -> 0.197 | Improved |
| add learnable prompts to DSPA (ablation) | BEVDet + DSPA · nuScenes val | rc-mAP: 0.164 -> 0.178 | Improved |
| add spatial/feature losses alongside channel loss in DFCD (ablation) | DFCD · nuScenes val | mAP: 0.429 -> 0.433 | Improved |
| add DFCD, VLCL, DSPA modules (full CLIPDet3D) vs GeoBEV | GeoBEV · nuScenes val rare categories | rc-mAP: 31.9 -> 35.2 | Improved |
| CLIPDet3D vs GeoBEV on Trailer class | GeoBEV · nuScenes val, Trailer | AP: 21.9 -> 17.5 | Hurt |
| higher BEV resolution and CLIPDet3D design vs BEVFormer | BEVFormer · nuScenes val | FPS: 3.9 -> 4.5 | Improved |
| CLIPDet3D vs CLIP-BEVFormer on R101 | CLIP-BEVFormer · nuScenes val (R101) | mAP: 0.467 -> 0.467 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- GeoBEV — GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (AAAI 2025) · better on 4 of 5 shared results
- BEVFormer — BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers (ECCV 2022) · better on 2 of 2 shared results
- SA-BEV — SA-BEV: Generating Semantic-Aware Bird's-Eye-View Feature for Multi-view 3D Object Detection (ICCV 2023) · 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
- Attention-Based Depth Distillation with 3D-Aware Positional Encoding for Monocular 3D Object Detection (AAAI 2023 · similarity 0.89)
- SemLT3D: Semantic-Guided Expert Distillation for Camera-only Long-Tailed 3D Object Detection (CVPR 2026 · similarity 0.89)
- Advancing 3D Object Detection With Depth-Aware Spatial Knowledge Distillation (TPAMI 2025 · similarity 0.89)
- DALDet: Depth-Aware Learning Based Object Detection for Autonomous Driving (AAAI 2024 · similarity 0.87)
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · similarity 0.87)
- PromptDet: A Lightweight 3D Object Detection Framework with LiDAR Prompts (AAAI 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.
