Research map / Detection and segmentation / 3D detection and driving
Multi-camera BEV detection
Detect 3D objects from multiple cameras with BEV representations and cross-view attention.
69 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in 3D detection and driving, keyword cluster “bev · autonomous · radar”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Few recent reports None of the tracked benchmarks has a report from 2025–26.
33 papers in 2023–24 and 29 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 20.9% to 19.1%.
Open questions the data points to
- Papers report BEVFusion* on nuScenes val (mAP) anywhere from 53.5 to 67.6: comparisons against it depend on whose number you take.
No longer reported by papers here: nuScenes validation (mAP, last in 2024); nuScenes validation (NDS, last in 2024).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2018 · Multi-View Harmonized Bilinear Network for 3D Object Recognition (CVPR 2018 · 334 citations)
- 2019 · Multi-Task Multi-Sensor Fusion for 3D Object Detection (CVPR 2019 · 779 citations)
- 2020 · Every View Counts: Cross-View Consistency in 3D Object Detection with Hybrid-Cylindrical-Spherical Voxelization (NeurIPS 2020 · 68 citations)
- 2022 · VISTA: Boosting 3D Object Detection via Dual Cross-VIew SpaTial Attention (CVPR 2022 · 99 citations)
- 2023 · BEVFormer v2: Adapting Modern Image Backbones to Bird's-Eye-View Recognition via Perspective Supervision (CVPR 2023 · 322 citations)
- 2024 · IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection (CVPR 2024 · 118 citations)
- 2025 · GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (AAAI 2025 · 20 citations)
- 2026 · SM3Det: A Unified Model for Multi-Modal Remote Sensing Object Detection (AAAI 2026 · 10 citations)
Earliest papers here
- Multi-View Harmonized Bilinear Network for 3D Object Recognition (CVPR 2018 · 334 citations)
- Multi-Task Multi-Sensor Fusion for 3D Object Detection (CVPR 2019 · 779 citations)
- Every View Counts: Cross-View Consistency in 3D Object Detection with Hybrid-Cylindrical-Spherical Voxelization (NeurIPS 2020 · 68 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
nuScenes validation · mAP (higher is better)
4 papers, 2022–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 60.81 | VISTA: Boosting 3D Object Detection via Dual Cross-VIew SpaTial Attention (CVPR 2022 · 99 citations) |
| 2024 | 72.8 | IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection (CVPR 2024 · 118 citations) |
nuScenes validation · NDS (higher is better)
4 papers, 2022–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 68.1 | VISTA: Boosting 3D Object Detection via Dual Cross-VIew SpaTial Attention (CVPR 2022 · 99 citations) |
| 2024 | 74 | IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection (CVPR 2024 · 118 citations) |
What has been tried
54 papers here report 680 measured changes (a component added, removed or swapped, compared with a baseline): helped 77%, no clear effect 4%, hurt 17%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add HSF to simple multimodal fusion | nuScenes validation · mAP | 69.4 -> 71.6 | Helped | IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection (CVPR 2024 · 118 citations) |
| Add CRF modulation | nuScenes val · NDS | 0.490 -> 0.492 | Helped | BEVNeXt: Reviving Dense BEV Frameworks for 3D Object Detection (CVPR 2024 · 59 citations) |
| Add additional downsampling and sparse height compression | nuScenes test · mAP | 68.9 -> 69.9 | Helped | GAFusion: Adaptive Fusing LiDAR and Camera with Multiple Guidance for 3D Object Detection (CVPR 2024 · 34 citations) |
| Add spread voxel pooling to BEVDepth | DAIR-V2X-I, vehicle (Middle) · AP3D|R40 | 60.37 -> 64.09 | Helped | BEVSpread: Spread Voxel Pooling for Bird's-Eye-View Representation in Vision-based Roadside 3D Object Detection (CVPR 2024 · 24 citations) |
| Distill from a fusion teacher using a simulated multi-modal student instead of a single-modal student | nuScenes validation · mAP | 35.94 -> 38.34 | Helped | SimDistill: Simulated Multi-Modal Distillation for BEV 3D Object Detection (AAAI 2024 · 23 citations) |
| Add GT-QI alongside GT-BEV guidance | nuScenes val · NDS | 37.3 -> 38.5 | Helped | CLIP-BEVFormer: Enhancing Multi-View Image-Based BEV Detector with Ground Truth Flow (CVPR 2024 · 23 citations) |
| Add RC-Sampling | nuScenes ablation · NDS | 0.456 -> 0.489 | Helped | GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection (AAAI 2025 · 20 citations) |
| Use binary occupancy labels instead of semantic labels | nuScenes test, ResNet-101 · NDS (%) | 58.1 -> 57.8 | Hurt | SOGDet: Semantic-Occupancy Guided Multi-View 3D Object Detection (AAAI 2024 · 16 citations) |
| add coordinate system transformation from Cartesian to polar coordinates | nuScenes validation · mean IoU | 21.6 -> 24.4 | Helped | Improving Bird's Eye View Semantic Segmentation by Task Decomposition (CVPR 2024 · 14 citations) |
| add SPM | nuScenes validation ablation · NDS | 0.425 -> 0.443 | Helped | Instance-Aware Multi-Camera 3D Object Detection with Structural Priors Mining and Self-Boosting Learning (AAAI 2024 · 12 citations) |
| add CMKI | nuScenes → Waymo · BEV AP / 3D AP | 39.18 / 20.78 -> 44.41 / 22.26 | Helped | CMDA: Cross-Modal and Domain Adversarial Adaptation for LiDAR-Based 3D Object Detection (AAAI 2024 · 11 citations) |
| add SAM instead of omitting spatial fusion | OPV2V · AP0.5 | 0.847 -> 0.940 | Helped | SparseAlign: a Fully Sparse Framework for Cooperative Object Detection (CVPR 2025 · 10 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| BEVFormer-tiny | nuScenes val | NDS | 35.5 – 39.02 | 3 papers, e.g. CVPR 2024 and CVPR 2023 |
| BEVFusion* | nuScenes val | mAP | 53.5 – 67.6 | 2 papers, e.g. AAAI 2025 and CVPR 2025 |
| BEVFusion* | nuScenes val | NDS | 60.4 – 70.9 | 2 papers, e.g. AAAI 2025 and CVPR 2025 |
| StreamPETR | nuScenes val | NDS | 0.54 – 0.609 | 2 papers, e.g. NeurIPS 2023 and CVPR 2024 |
What reviewers will expect
Datasets and baselines named in the 15 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- nuScenes · 11 of 15 papers
- SemanticKITTI · 2 of 15 papers
Baselines
- BEVDepth · 4 of 15 papers
- BEVDet · 3 of 15 papers
- BEVFormer · 3 of 15 papers
- BEVStereo · 2 of 15 papers
- RayDN · 2 of 15 papers
- StreamPETR · 2 of 15 papers
- SOLOFusion · 2 of 15 papers
- BEVFusion · 2 of 15 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 62 papers; from OpenAlex and the papers' affiliations).
- Fudan University · 6 papers
- Shanghai Jiao Tong University · 6 papers
- Chinese Academy of Sciences · 5 papers
- University of Chinese Academy of Sciences · 4 papers
- Shanghai AI Laboratory · 4 papers
- The University of Sydney · 3 papers
- University of Science and Technology of China · 3 papers
- NVIDIA · 3 papers
- Zhejiang University · 3 papers
- Samsung · 2 papers
Use it
Download BibTeX of 37 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- LiDAR perception for driving (75 papers)
- Camera-radar fusion detection (36 papers)
Elsewhere in the topic
- Unified LiDAR 3D detection (107 papers)
- Depth-based and label-free 3D detection (58 papers)
- 3D objectness and scene segmentation (57 papers)
- Monocular 3D object detection (56 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).