Research map / Detection and segmentation / 3D detection and driving
C2E: Boosting Ego-Only 3D Object Detection via Multi-Teacher Contrastive Knowledge Distillation
Jinlong Wang, Xun Huang, Qiming Xia, Shijia Zhao, 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 |
|---|---|---|---|
| distill from a single collaborative teacher (AttFuse) into the ego-only student (ablation) | M2S student · V2XSet test | 3D mAP@0.7: 58.06 -> 58.29 | No clear effect |
| replace single-teacher with multi-teacher adaptive distillation (ablation) | M2S student · V2XSet test | 3D mAP@0.7: 58.29 -> 62.76 | Improved |
| add multi-level feature enhancement (MLFE) (ablation) | M2S student · V2XSet test | 3D mAP@0.7: 62.76 -> 64.58 | Improved |
| add auxiliary point cloud reconstruction (APCR) (ablation) | M2S student · V2XSet test | 3D mAP@0.7: 64.58 -> 66.50 | Improved |
| add multi-teacher contrastive learning (local distillation) (ablation) | M2S student · V2XSet test | 3D mAP@0.7: 66.50 -> 66.70; AP@0.5 79.32 -> 79.26 | Mixed |
| use three teachers instead of one or two (ablation) | M2S student · V2XSet val | AP@0.7: 73.26 -> 73.73 | Improved |
| apply M2S distillation to six detectors with ego-only input | AttFuse, V2X-ViT, Where2comm, CoAlign, CoSDH, PillarNeSt · V2XSet test | 3D mAP@0.7: 58.06 -> 66.70 (AttFuse, +8.64) | Improved |
| apply M2S distillation on real-world V2V4Real at strict IoU | Where2comm, CoSDH · V2V4Real val | AP@0.7: 27.03 -> 25.55 (Where2comm); -0.65 (CoSDH) | Mixed |
| apply M2S on DAIR-V2X | AttFuse · DAIR-V2X val | AP@0.7: 33.25 -> 39.86 | Improved |
| apply M2S at far range | AttFuse · V2XSet val, range [50,+inf) m | AP@0.7: 39.63 -> 39.97 | No clear effect |
Closest papers by meaning
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · similarity 0.88)
- CoLC: Communication-Efficient Collaborative Perception with LiDAR Completion (CVPR 2026 · similarity 0.88)
- INSTINCT: Instance-Level Interaction Architecture for Query-Based Collaborative Perception (ICCV 2025 · similarity 0.88)
- Learning to Detect Objects from Multi-Agent LiDAR Scans without Manual Labels (CVPR 2025 · similarity 0.88)
- CoGoal3D: Collaborative 3D Object Detection with 3D-Aware Fusion and Refinement (ECCV 2026 · similarity 0.88)
- CALICO: Self-Supervised Camera-LiDAR Contrastive Pre-training for BEV Perception (ICLR 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.
