Research map / Detection and segmentation / 3D detection and driving
Horizon3D: Sparse Radar-Camera Fusion for Long-Range 3D Perception in Autonomous Driving
Geonho Bang, Geunju Baek, DongYoung Lee, Wonjun Jeong, 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 OCSF with random Gaussians to radar-only baseline (ablation) | radar-only · TruckScenes val (ResNet-50) | mAP: 12.8 -> 17.7 | Improved |
| replace random Gaussian init with KGGI (ablation) | OCSF · TruckScenes val (ResNet-50) | mAP: 17.7 -> 20.3 | Improved |
| add DPTF without KGGI (ablation) | OCSF-only · TruckScenes val (ResNet-50) | mAP: 17.7 -> 23.0 | Improved |
| add DPTF to KGGI+OCSF (ablation) | Horizon3D single-frame · TruckScenes val (ResNet-50) | NDS: 35.4 -> 37.4 | Improved |
| keypoint-only init instead of random init (ablation) | KGGI · TruckScenes val (ResNet-50) | mAP: 17.7 -> 19.2 | Improved |
| add random Gaussians to keypoint-only init (ablation) | KGGI · TruckScenes val (ResNet-50) | mAP: 19.2 -> 20.3 | Improved |
| enlarge GT boxes for occupancy supervision (ablation) | OCSF · TruckScenes val (ResNet-50) | mAP: 17.2 -> 20.3 | Improved |
| enlargement margin 4.8 m vs 0.8 m or 6.4 m (ablation) | OCSF · TruckScenes val (ResNet-50) | mAP: 18.6 -> 20.3 (6.4 m: 19.6) | Mixed |
| add Gaussian path to BEV path (ablation) | DPTF · TruckScenes val (ResNet-50) | mAP: 22.5 -> 23.6 | Improved |
| add object velocity compensation to BEV path (ablation) | DPTF BEV path · TruckScenes val (ResNet-50) | mAP: 21.6 -> 23.0 | Improved |
| add object velocity compensation to Gaussian path (ablation) | DPTF Gaussian path · TruckScenes val (ResNet-50) | NDS: 34.8 -> 36.0 | Improved |
| FPS instead of top-K or random Gaussian sampling (ablation) | Gaussian path · TruckScenes val (ResNet-50) | mAP: 21.1 -> 21.3 (random 20.7) | Improved |
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.
- RCTrans — RCTrans: Radar-Camera Transformer via Radar Densifier and Sequential Decoder for 3D Object Detection (AAAI 2025) · better on 2 of 2 shared results
- Far3D — Far3D: Expanding the Horizon for Surround-View 3D Object Detection (AAAI 2024) · better on 2 of 2 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
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
- CRT-Fusion: Camera, Radar, Temporal Fusion Using Motion Information for 3D Object Detection (NeurIPS 2024 · similarity 0.91)
- RaCFormer: Towards High-Quality 3D Object Detection via Query-based Radar-Camera Fusion (CVPR 2025 · similarity 0.90)
- RCBEVDet: Radar-camera Fusion in Bird's Eye View for 3D Object Detection (CVPR 2024 · similarity 0.89)
- HGSFusion: Radar-Camera Fusion with Hybrid Generation and Synchronization for 3D Object Detection (AAAI 2025 · similarity 0.88)
- IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection (CVPR 2024 · similarity 0.88)
- RadarOcc: Robust 3D Occupancy Prediction with 4D Imaging Radar (NeurIPS 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.
