Research map / Detection and segmentation / 3D detection and driving
SpatialDETR: Robust Scalable Transformer-Based 3D Object Detection from Multi-View Camera Images with Global Cross-Sensor Attention
Simon Doll, Richard Schulz, Lukas Schneider, Viviane Benzin, Markus Enzweiler, Hendrik P.A. Lensch
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 sensor-relative query projection P(Q) (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.248 -> 0.317 | Improved |
| add value projection P(V) to P(Q) (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.317 -> 0.313 | Hurt |
| add query-center visibility constraint C(Q) to P(Q) (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.317 -> 0.318 | Mixed |
| add C(Q) to P(Q)+P(V) (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.313 -> 0.319 | Improved |
| increase object queries from 600 to 900 (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.308 -> 0.313 | Improved |
| increase object queries from 900 to 1200 (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.313 -> 0.313 | No clear effect |
| increase decoder layers from 2 to 6 (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.308 -> 0.313 | Improved |
| increase decoder layers from 6 to 10 (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.313 -> 0.311 | No clear effect |
| share weights across decoder layers 2-6 (ablation) | SpatialDETR · nuScenes val (ablation setup) | mAP: 0.313 -> 0.314 | Improved |
| initialize backbone from FCOS3D checkpoint (PT) | SpatialDETR · nuScenes val | mAP: 0.303 -> 0.351 | Improved |
| use spatially-aware global attention instead of DETR3D center lookup | truck/bus AP · nuScenes val | AP truck: 0.286 -> 0.302 | Improved |
| use spatially-aware global attention instead of DETR3D | bus AP · nuScenes val | AP bus: 0.347 -> 0.378 | Improved |
Closest papers by meaning
- Viewpoint Equivariance for Multi-View 3D Object Detection (CVPR 2023 · similarity 0.89)
- DESTR: Object Detection With Split Transformer (CVPR 2022 · similarity 0.88)
- Spatial-Temporal Graph Enhanced DETR Towards Multi-Frame 3D Object Detection (TPAMI 2024 · similarity 0.88)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.87)
- MonoDGP: Monocular 3D Object Detection with Decoupled-Query and Geometry-Error Priors (CVPR 2025 · similarity 0.86)
- Cubify Anything: Scaling Indoor 3D Object Detection (CVPR 2025 · similarity 0.86)
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.
