Research map / Detection and segmentation / 3D detection and driving
OcRFDet: Object-Centric Radiance Fields for Multi-View 3D Object Detection in Autonomous Driving
Mingqian Ji, Shanshan Zhang, Jian Yang
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 scene-level radiance field rendering as auxiliary task (ablation) | DualBEV · nuScenes val | mAP: 34.1 -> 33.8 | Hurt |
| restrict rendering loss to foreground objects (object-centric) (ablation) | DualBEV · nuScenes val | mAP: 33.8 -> 34.7 | Improved |
| add scene-level warm-up before object-centric optimization (ablation) | OcRF · nuScenes val | mAP: 34.7 -> 35.0 | Improved |
| add OcRF auxiliary rendering (ablation) | DualBEV · nuScenes val | mAP: 34.1 -> 35.0 | Improved |
| add height-aware opacity-based attention (HOA) on top of OcRF (ablation) | DualBEV + OcRF · nuScenes val | mAP: 35.0 -> 35.5 | Improved |
| use hybrid 3DGS plus NeRF instead of a single radiance field (ablation) | OcRF · nuScenes val | mAP: 34.7 -> 35.0 | Improved |
| add depth map rendering besides RGB (ablation) | OcRF · nuScenes val | mAP: 34.6 -> 35.0 | Improved |
| fuse 3DGS and NeRF opacity instead of a single source (ablation) | HOA · nuScenes val | mAP: 34.7 -> 35.5 | Improved |
| use height slice attention with multiple parallel groups (k=4) instead of none or shared weights (ablation) | HOA · nuScenes val | mAP: 34.9 -> 35.5 | Improved |
| increase height groups k from 4 to 6 (ablation) | HSA · nuScenes val | mAP: 35.5 -> 35.5 | No clear effect |
| render all six views instead of one random view (ablation) | OcRF · nuScenes val | mAP: 35.0 -> 34.5 | Hurt |
| cross-attention opacity fusion instead of weighted mean or concat-conv (ablation) | HOA · nuScenes val | mAP: 35.3 / 35.0 -> 35.5 | Improved |
Closest papers by meaning
- GO-N3RDet: Geometry Optimized NeRF-enhanced 3D Object Detector (CVPR 2025 · similarity 0.89)
- OpenAD: Open-World Autonomous Driving Benchmark for 3D Object Detection (NeurIPS 2025 · similarity 0.89)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 2025 · similarity 0.88)
- Towards Flexible 3D Perception: Object-Centric Occupancy Completion Augments 3D Object Detection (NeurIPS 2024 · similarity 0.88)
- SOGDet: Semantic-Occupancy Guided Multi-View 3D Object Detection (AAAI 2024 · similarity 0.87)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 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.
