Research map / Detection and segmentation / 3D detection and driving
X3KD: Knowledge Distillation Across Modalities, Tasks and Stages for Multi-Camera 3D Object Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 X-OD, X-FD, X-AT, and X-IS (ablation) | BEVDepth† camera detector · nuScenes validation | mAP: 35.9 -> 39.0 | Helped |
| Add X-OD, X-FD, X-AT, and X-IS (ablation) | BEVDepth† camera detector · nuScenes validation | NDS: 47.2 -> 50.5 | Helped |
| Add output distillation (X-OD) alone (ablation) | BEVDepth† · nuScenes validation | mAP: 35.9 -> 35.7 | Hurt |
| Add output distillation (X-OD) alone (ablation) | BEVDepth† · nuScenes validation | NDS: 47.2 -> 48.7 | Helped |
| Add feature distillation (X-FD) alone (ablation) | BEVDepth† · nuScenes validation | mAP: 35.9 -> 36.1 | Helped |
| Add adversarial feature training (X-AT) alone (ablation) | BEVDepth† · nuScenes validation | mAP: 35.9 -> 35.5 | Hurt |
| Add X-OD, X-FD, and X-AT together (ablation) | BEVDepth† · nuScenes validation | NDS: 47.2 -> 49.4 | Helped |
| Add instance segmentation distillation (X-IS) (ablation) | BEVDepth† · nuScenes validation | mAP: 35.9 -> 38.7 | Helped |
| Weight X-OD regression loss by teacher output probabilities (ablation) | X-OD camera detector · nuScenes validation | mAOE: 1.090 -> 0.456 | Helped |
| Train X-OD without ground-truth annotations (ablation) | BEVDepth† camera detector · nuScenes validation | mAP: 35.9 -> 36.5 | Helped |
| Initialize the backbone with instance-segmentation pretraining (ablation) | BEVDepth† camera detector · nuScenes validation | mAP: 35.9 -> 37.7 | Helped |
| Add X-IS distillation to the pretrained camera detector (ablation) | ResNet-50 camera detector · nuScenes validation | mAP: 37.7 -> 38.7 | Helped |
Closest papers by meaning
- SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection (AAAI 2024 · similarity 0.89)
- Weak-to-Strong 3D Object Detection with X-Ray Distillation (CVPR 2024 · similarity 0.88)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.88)
- CMDA: Cross-Modal and Domain Adversarial Adaptation for LiDAR-Based 3D Object Detection (AAAI 2024 · similarity 0.87)
- PromptDet: A Lightweight 3D Object Detection Framework with LiDAR Prompts (AAAI 2025 · similarity 0.87)
- CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation (CVPR 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.