Research map / Detection and segmentation / 3D detection and driving
SimPB: A Single Model for 2D and 3D Object Detection from Multiple Cameras
Yingqi Tang, Zhaotie Meng, Guoliang Chen, Erkang Cheng
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 a multi-view 2D decoder layer interleaved with 3D layers (cyclic 3D-2D-3D) (ablation) | SimPB hybrid decoder · nuScenes val 3D (24 epochs) | NDS: 0.504 -> 0.527 | Improved |
| cyclic hybrid layers instead of cascaded 2D then 3D layers (two-stage style) (ablation) | SimPB · nuScenes val 3D (24 epochs) | NDS: 0.523 -> 0.527 | Improved |
| dynamic camera-parameter allocation (object center + anchor corners) instead of uniform allocation (ablation) | SimPB · nuScenes val 3D (24 epochs) | mAP: 0.365 -> 0.421 | Improved |
| query fusion instead of cross-attention for aggregating 2D into 3D queries (ablation) | SimPB · nuScenes val 3D (24 epochs) | mAP: 0.399 -> 0.421 | Improved |
| add 3D deep supervision on aggregated 3D queries (ablation) | SimPB query fusion · nuScenes val 3D (24 epochs) | NDS: 0.489 -> 0.527 | Improved |
| add observation angle loss (lambda 0.5) (ablation) | SimPB · nuScenes val 3D (24 epochs) | mAOE: 0.611 -> 0.492 | Improved |
| increase lambda_alpha from 0.5 to 1.00 (ablation) | SimPB · nuScenes val 3D (24 epochs) | mAP: 0.421 -> 0.418 | No clear effect |
| add one encoder layer (ablation) | SimPB · nuScenes val 3D (24 epochs) | mAP: 0.412 -> 0.421 | Improved |
| increase encoder layers from 1 to 4 (ablation) | SimPB · nuScenes val 3D (24 epochs) | mAP / FPS: mAP 0.421 -> 0.439, FPS 10.9 -> 6.1 | Mixed |
| plug SimPB hybrid decoder into DETR3D | DETR3D · nuScenes val 3D | NDS: 0.306 -> 0.358 | Improved |
| plug SimPB hybrid decoder into Sparse4Dv2 | Sparse4Dv2 · nuScenes val 3D | mAP: 0.367 -> 0.398 | Improved |
| use nuImages perspective-pretrained backbone | SimPB ResNet50 · nuScenes val 3D | mAP: 0.475 -> 0.487 | Improved |
Later papers that compared with this method
1 later accepted paper reports SimPB as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Rethinking the Spatio-Temporal Alignment of End-to-End 3D Perception (AAAI 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DeformableDETR — Deformable DETR: Deformable Transformers for End-to-End Object Detection (ICLR 2021) · 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
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.87)
- UniMODE: Unified Monocular 3D Object Detection (CVPR 2024 · similarity 0.87)
- GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (AAAI 2025 · similarity 0.87)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 2025 · similarity 0.87)
- Instance-Aware Multi-Camera 3D Object Detection with Structural Priors Mining and Self-Boosting Learning (AAAI 2024 · similarity 0.86)
- BEVNeXt: Reviving Dense BEV Frameworks for 3D Object Detection (CVPR 2024 · 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.
