Research map / Detection and segmentation / 3D detection and driving
RCTDistill: Cross-Modal Knowledge Distillation Framework for Radar-Camera 3D Object Detection with Temporal Fusion
Geonho Bang, Minjae Seong, Jisong Kim, Geunju Baek, Daye Oh, Junhyung Kim, Junho Koh, Jun Won Choi
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 RAKD (elliptical range-azimuth mask) alone (ablation) | BEVFusion-R temporal student · nuScenes val (ablation) | NDS: 55.0 -> 57.2 | Improved |
| add TKD alone (ablation) | BEVFusion-R temporal student · nuScenes val (ablation) | mAP: 47.1 -> 48.5 | Improved |
| add RDKD alone (ablation) | BEVFusion-R temporal student · nuScenes val (ablation) | mAP: 47.1 -> 49.0 | Improved |
| combine RAKD, TKD and RDKD together (ablation) | RCTDistill · nuScenes val (ablation) | NDS: 57.6 -> 58.4 | Improved |
| use elliptical Gaussian mask instead of BEVDistill circular mask (ablation) | RAKD · nuScenes val (ablation) | mAP: 48.3 -> 49.3 | Improved |
| use proposed TKD instead of STXD (ablation) | temporal KD · nuScenes val (ablation) | NDS: 56.5 -> 57.0 | Improved |
| use proposed RDKD instead of CRKD affinity distillation (ablation) | relation KD · nuScenes val (ablation) | mAP: 48.1 -> 49.0 | Improved |
| apply the full distillation framework to the baseline | BEVFusion-R student (R50) · nuScenes val (R50) | NDS: 57.3 -> 62.2 | Improved |
| apply distillation to the baseline | BEVFusion-R student (R50) · nuScenes val (R50) | FPS: 27.8 -> 26.2 | Mixed |
| apply the full distillation framework to the baseline | BEVFusion-R student (R101) · nuScenes val (R101) | mAP: 54.2 -> 59.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CRT-Fusion — CRT-Fusion: Camera, Radar, Temporal Fusion Using Motion Information for 3D Object Detection (NeurIPS 2024) · better on 2 of 2 shared results
- CRKD — CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation (CVPR 2024) · 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
- CRKD: Enhanced Camera-Radar Object Detection with Cross-modality Knowledge Distillation (CVPR 2024 · similarity 0.93)
- RadarDistill: Boosting Radar-based Object Detection Performance via Knowledge Distillation from LiDAR Features (CVPR 2024 · similarity 0.92)
- SCKD: Semi-Supervised Cross-Modality Knowledge Distillation for 4D Radar Object Detection (AAAI 2025 · similarity 0.90)
- CRT-Fusion: Camera, Radar, Temporal Fusion Using Motion Information for 3D Object Detection (NeurIPS 2024 · similarity 0.89)
- RCBEVDet: Radar-camera Fusion in Bird's Eye View for 3D Object Detection (CVPR 2024 · similarity 0.88)
- SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection (AAAI 2024 · similarity 0.88)
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.
