Research map / Detection and segmentation / 3D detection and driving
Unified LiDAR 3D detection
Detect 3D objects from point clouds with grouping, graphs, and domain adaptation.
107 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in 3D detection and driving, keyword cluster “3d object · monocular · kitti”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Few recent reports The tracked benchmarks have at most one report from 2025–26 that could beat the old record, too few to say whether they are still moving.
31 papers in 2023–24 and 32 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 19.6% to 21.1%.
Open questions the data points to
- Papers report VoteNet* on ScanNet V2 (mAP@0.5) anywhere from 28.3 to 44.2: comparisons against it depend on whose number you take.
- Papers report VoteNet* on ScanNet V2 (mAP@0.25) anywhere from 50.4 to 63.8: comparisons against it depend on whose number you take.
No longer reported by papers here: ScanNet V2 (mAP@0.5, last in 2022).
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
- 2016 · Deep Sliding Shapes for Amodal 3D Object Detection in RGB-D Images (CVPR 2016 · 726 citations)
- 2017 · 3D Bounding Box Estimation Using Deep Learning and Geometry (CVPR 2017 · 1,208 citations)
- 2018 · 3D Object Detection With Latent Support Surfaces (CVPR 2018 · 33 citations)
- 2019 · Stereo R-CNN Based 3D Object Detection for Autonomous Driving (CVPR 2019 · 640 citations)
- 2020 · MLCVNet: Multi-Level Context VoteNet for 3D Object Detection (CVPR 2020 · 213 citations)
- 2021 · Center-Based 3D Object Detection and Tracking (CVPR 2021 · 2,039 citations)
- 2022 · Focal Sparse Convolutional Networks for 3D Object Detection (CVPR 2022 · 320 citations)
- 2023 · VoxelNeXt: Fully Sparse VoxelNet for 3D Object Detection and Tracking (CVPR 2023 · 464 citations)
- 2024 · Far3D: Expanding the Horizon for Surround-View 3D Object Detection (AAAI 2024 · 73 citations)
- 2025 · UniDet3D: Multi-dataset Indoor 3D Object Detection (AAAI 2025 · 19 citations)
- 2026 · WinMamba: Multi-Scale Shifted Windows in State Space Model for 3D Object Detection (AAAI 2026 · 3 citations)
Earliest papers here
- Deep Sliding Shapes for Amodal 3D Object Detection in RGB-D Images (CVPR 2016 · 726 citations)
- Deep Exemplar 2D-3D Detection by Adapting From Real to Rendered Views (CVPR 2016 · 106 citations)
- Motion From Structure (MfS): Searching for 3D Objects in Cluttered Point Trajectories (CVPR 2016 · 3 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 test · NDS (higher is better)
9 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 65.5 | Center-Based 3D Object Detection and Tracking (CVPR 2021 · 2,039 citations) |
| 2022 | 76.3 | DeepInteraction: 3D Object Detection via Modality Interaction (NeurIPS 2022 · 63 citations) |
nuScenes test · mAP (higher is better)
9 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 58 | Center-Based 3D Object Detection and Tracking (CVPR 2021 · 2,039 citations) |
| 2022 | 75.6 | DeepInteraction: 3D Object Detection via Modality Interaction (NeurIPS 2022 · 63 citations) |
SUN RGB-D · mAP@0.25 (higher is better)
6 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 59.9 | Semantic Consistency Networks for 3D Object Detection (AAAI 2021 · 8 citations) |
| 2022 | 66.8 | CAGroup3D: Class-Aware Grouping for 3D Object Detection on Point Clouds (NeurIPS 2022 · 37 citations) |
ScanNet V2 · mAP@0.25 (higher is better)
6 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 63.3 | Semantic Consistency Networks for 3D Object Detection (AAAI 2021 · 8 citations) |
| 2022 | 75.1 | CAGroup3D: Class-Aware Grouping for 3D Object Detection on Point Clouds (NeurIPS 2022 · 37 citations) |
nuScenes val · NDS (higher is better)
6 papers, 2022–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 72.6 | DeepInteraction: 3D Object Detection via Modality Interaction (NeurIPS 2022 · 63 citations) |
ScanNet V2 · mAP@0.5 (higher is better)
5 papers, 2021–2022 · not reported since 2022
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 40.5 | Semantic Consistency Networks for 3D Object Detection (AAAI 2021 · 8 citations) |
| 2022 | 61.3 | CAGroup3D: Class-Aware Grouping for 3D Object Detection on Point Clouds (NeurIPS 2022 · 37 citations) |
What has been tried
87 papers here report 1,003 measured changes (a component added, removed or swapped, compared with a baseline): helped 80%, no clear effect 3%, hurt 15%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add geometry-aware voting | ScanNet V2 val · mAP@0.25 | 68.2 to 71.3 | Helped | SPGroup3D: Superpoint Grouping Network for Indoor 3D Object Detection (AAAI 2024 · 22 citations) |
| Add decoupled pseudo-label generation (DPG) | KITTI validation, Car · AP3D|R40 Mod. | 18.02 -> 19.23 | Helped | Decoupled Pseudo-labeling for Semi-Supervised Monocular 3D Object Detection (CVPR 2024 · 21 citations) |
| jointly train across datasets rather than train separately | S3DIS · mAP50 | 39.7 -> 57.9 | Helped | UniDet3D: Multi-dataset Indoor 3D Object Detection (AAAI 2025 · 19 citations) |
| add semantic ControlNet | Omni3D-SUNRGBD, tuned · AP3D | 23.11 -> 27.81 | Helped | 3DiffTection: 3D Object Detection with Geometry-Aware Diffusion Features (CVPR 2024 · 14 citations) |
| add SIE | Real3D-AD · I-AUROC | 0.656 -> 0.757 | Helped | Look Inside for More: Internal Spatial Modality Perception for 3D Anomaly Detection (AAAI 2025 · 12 citations) |
| use Faster-RCNN instead of DAB-DETR as the 2D detector | nuScenes val · NDS | 50.2 -> 50.0 | No clear effect | Enhancing 3D Object Detection with 2D Detection-Guided Query Anchors (CVPR 2024 · 12 citations) |
| Use high-resolution FARO-derived depth rather than onboard LiDAR depth | CA-1M · AP25 | 40.9 -> 46.7 | Helped | Cubify Anything: Scaling Indoor 3D Object Detection (CVPR 2025 · 10 citations) |
| Add the source-domain intra-domain triplet loss | N → K (Car) · AP3D | 57.79 -> 58.21 | Helped | Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D Object Detection (CVPR 2024 · 10 citations) |
| Use shift-window context instead of global cluster context | nuScenes val · NDS | 51.7 -> 51.5 (-0.2) | Hurt | Multi-View Attentive Contextualization for Multi-View 3D Object Detection (CVPR 2024 · 9 citations) |
| Add intensity histogram input | KITTI val, Pedestrian 3DR40 · mAP | 49.62 -> 51.07 | Helped | PillarHist: A Quantization-aware Pillar Feature Encoder based on Height-aware Histogram (CVPR 2025 · 8 citations) |
| add PLM using classification-score label matching | ScanNet V2 one object per scene · mAP@0.25 | 37.6 to 38.0 | Helped | Learning Class Prototypes for Unified Sparse-Supervised 3D Object Detection (CVPR 2025 · 7 citations) |
| Add DCPG to mask shrink | KITTI val split, Car-3D AP@IoU 0.7 · AP | 35.10 -> 38.26; 23.75 -> 29.25; 19.52 -> 24.74 | Helped | SP3D: Boosting Sparsely-Supervised 3D Object Detection via Accurate Cross-Modal Semantic Prompts (CVPR 2025 · 7 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 |
|---|---|---|---|---|
| VoteNet* | ScanNet V2 | mAP@0.5 | 28.3 – 44.2 | 2 papers, e.g. CVPR 2022 and CVPR 2022 |
| VoteNet* | ScanNet V2 | mAP@0.25 | 50.4 – 63.8 | 2 papers, e.g. CVPR 2022 and CVPR 2022 |
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
- ScanNet · 5 of 15 papers
- SUN RGB-D · 4 of 15 papers
- nuScenes · 4 of 15 papers
- KITTI · 4 of 15 papers
- ARKitScenes · 2 of 15 papers
- ScanNet V2 · 2 of 15 papers
- Waymo · 2 of 15 papers
Baselines
- CenterPoint · 3 of 15 papers
- 3DETR · 2 of 15 papers
- FCAF3D · 2 of 15 papers
- TR3D · 2 of 15 papers
- ImVoxelNet · 2 of 15 papers
- VoxelNeXt · 2 of 15 papers
- CoIn · 2 of 15 papers
- SS3D · 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 63 papers; from OpenAlex and the papers' affiliations).
- University of Oxford · 3 papers
- Xiamen University · 2 papers
- Texas A&M University · 2 papers
- Megvii · 2 papers
- Beijing Institute of Technology · 2 papers
- Northwestern Polytechnical University · 2 papers
- Shanghai AI Laboratory · 2 papers
- Zhejiang University · 2 papers
- City University of Hong Kong · 2 papers
- Huawei · 2 papers
Use it
Download BibTeX of 40 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
- Depth-based and label-free 3D detection (58 papers)
- 3D objectness and scene segmentation (57 papers)
- Monocular 3D object detection (56 papers)
Elsewhere in the topic
- LiDAR perception for driving (75 papers)
- Multi-camera BEV detection (69 papers)
- Camera-radar fusion detection (36 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).