Research map / 3D vision / Human pose estimation
Transformer-based multi-view 3D pose
Uses transformers and diffusion with multi-view or occlusion-aware refinement for 3D human pose.
83 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in Human pose estimation, keyword cluster “human pose · pose estimation · 3d human”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 2 of 6 tracked benchmarks.
26 papers in 2023–24 and 34 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 9.6% to 11.3%.
Open questions the data points to
- MPI-INF-3DHP (PCK): best went from 98.5 (up to 2024) to 99.9 in 2025 — SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · 4 citations)
- MPI-INF-3DHP (AUC): best went from 78.8 (up to 2024) to 91.6 in 2025 — SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · 4 citations)
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 · Through-Wall Human Pose Estimation Using Radio Signals (CVPR 2018 · 679 citations)
- 2019 · PifPaf: Composite Fields for Human Pose Estimation (CVPR 2019 · 557 citations)
- 2020 · Epipolar Transformers (CVPR 2020 · 191 citations)
- 2021 · SimPoE: Simulated Character Control for 3D Human Pose Estimation (CVPR 2021 · 129 citations)
- 2022 · MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation (CVPR 2022 · 453 citations)
- 2023 · PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation (CVPR 2023 · 254 citations)
- 2024 · KTPFormer: Kinematics and Trajectory Prior Knowledge-Enhanced Transformer for 3D Human Pose Estimation (CVPR 2024 · 107 citations)
- 2025 · PoseMamba: Monocular 3D Human Pose Estimation with Bidirectional Global-Local Spatio-Temporal State Space Model (AAAI 2025 · 29 citations)
- 2026 · FMPose3D: monocular 3D pose estimation via flow matching (CVPR 2026 · 1 citation)
Earliest papers here
- Through-Wall Human Pose Estimation Using Radio Signals (CVPR 2018 · 679 citations)
- 3D Human Sensing, Action and Emotion Recognition in Robot Assisted Therapy of Children With Autism (CVPR 2018 · 108 citations)
- PifPaf: Composite Fields for Human Pose Estimation (CVPR 2019 · 557 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.
Human3.6M · MPJPE (lower is better)
5 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 56.7 | SimPoE: Simulated Character Control for 3D Human Pose Estimation (CVPR 2021 · 129 citations) |
| 2024 | 7.9 | FusionFormer: A Concise Unified Feature Fusion Transformer for 3D Pose Estimation (AAAI 2024 · 21 citations) |
MPI-INF-3DHP · PCK (higher is better)
5 papers, 2022–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 93.8 | MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation (CVPR 2022 · 453 citations) |
| 2023 | 97.9 | PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation (CVPR 2023 · 254 citations) |
| 2024 | 98.5 | Disentangled Diffusion-Based 3D Human Pose Estimation with Hierarchical Spatial and Temporal Denoiser (AAAI 2024 · 36 citations) |
| 2025 | 99.9 | SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · 4 citations) |
MPI-INF-3DHP · AUC (higher is better)
5 papers, 2022–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 63.3 | MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation (CVPR 2022 · 453 citations) |
| 2023 | 78.8 | PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation (CVPR 2023 · 254 citations) |
| 2025 | 91.6 | SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · 4 citations) |
3DPW · PA-MPJPE (lower is better)
4 papers, 2023–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 40.6 | NIKI: Neural Inverse Kinematics With Invertible Neural Networks for 3D Human Pose and Shape Estimation (CVPR 2023 · 75 citations) |
3DPW · MPJPE (lower is better)
4 papers, 2023–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 71.3 | NIKI: Neural Inverse Kinematics With Invertible Neural Networks for 3D Human Pose and Shape Estimation (CVPR 2023 · 75 citations) |
MPI-INF-3DHP · MPJPE (lower is better)
4 papers, 2023–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 27.8 | PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation (CVPR 2023 · 254 citations) |
| 2024 | 5.4 | FusionFormer: A Concise Unified Feature Fusion Transformer for 3D Pose Estimation (AAAI 2024 · 21 citations) |
What has been tried
60 papers here report 686 measured changes (a component added, removed or swapped, compared with a baseline): helped 78%, no clear effect 4%, hurt 17%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add KPA | Human3.6M (GT 2D, T=243) · MPJPE (mm) | 21.8 -> 20.0 | Helped | KTPFormer: Kinematics and Trajectory Prior Knowledge-Enhanced Transformer for 3D Human Pose Estimation (CVPR 2024 · 107 citations) |
| Use manually designed action-class prompts instead of no prompts | Human3.6M (DET) · MPJPE | 37.2 -> 35.8 | Helped | FinePOSE: Fine-Grained Prompt-Driven 3D Human Pose Estimation via Diffusion Models (CVPR 2024 · 55 citations) |
| Add disentangled input | Human3.6M · MPJPE | 40.23 -> 39.65 | Helped | Disentangled Diffusion-Based 3D Human Pose Estimation with Hierarchical Spatial and Temporal Denoiser (AAAI 2024 · 36 citations) |
| add bidirectional global-local spatial-temporal scanning instead of bidirectional scanning without local scan | Human3.6M · MPJPE | 42.4 -> 41.8 | Helped | PoseMamba: Monocular 3D Human Pose Estimation with Bidirectional Global-Local Spatio-Temporal State Space Model (AAAI 2025 · 29 citations) |
| increase the number of frames from T=1 to T=9 | Human3.6M (2DGT) · MPJPE | 8.87 -> 7.32 | Helped | FusionFormer: A Concise Unified Feature Fusion Transformer for 3D Pose Estimation (AAAI 2024 · 21 citations) |
| Increase initialized query count from 128 to 1024 | CMU Panoptic ablation · AP25 | 77.9 -> 92.3 | Helped | Multiple View Geometry Transformers for 3D Human Pose Estimation (CVPR 2024 · 19 citations) |
| Add Vertex-guided Adaptive Distillation | FreeMotion · SUCD | 3.17 -> 2.85 | Helped | LiveHPS: LiDAR-based Scene-level Human Pose and Shape Estimation in Free Environment (CVPR 2024 · 18 citations) |
| Add depth information to the CNN-based 3D module | UnrealEgo-RW · MPJPE | 115.36 -> 120.39 | Hurt | 3D Human Pose Perception from Egocentric Stereo Videos (CVPR 2024 · 17 citations) |
| Add Stage 2 focused refinement after the coarse pose | Human3.6M and MPI-INF-3DHP · MPJPE | 29.2 -> 28.6; 70.9 -> 68.9 | Helped | Lifting by Image – Leveraging Image Cues for Accurate 3D Human Pose Estimation (AAAI 2024 · 16 citations) |
| add hierarchical codebooks instead of one shared codebook | Human3.6M · MPJPE | 45.3 -> 42.0 | Helped | HiPART: Hierarchical Pose AutoRegressive Transformer for Occluded 3D Human Pose Estimation (CVPR 2025 · 12 citations) |
| Add spatial and edge feature embeddings to position features | Human3.6M · P1 (mm) | 29.4 -> 27.6 | Helped | Deep Semantic Graph Transformer for Multi-View 3D Human Pose Estimation (AAAI 2024 · 12 citations) |
| replace Grid ViT with channel-wise ViT encoder | UnrealEgo test · MPJPE | 49.03 -> 61.62 | Hurt | Attention-Propagation Network for Egocentric Heatmap to 3D Pose Lifting (CVPR 2024 · 9 citations) |
What reviewers will expect
Datasets and baselines named in the 16 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Human3.6M · 7 of 16 papers
- MPI-INF-3DHP · 4 of 16 papers
- 3DPW · 2 of 16 papers
Baselines
- MHFormer · 4 of 16 papers
- MixSTE · 3 of 16 papers
- MotionBERT · 2 of 16 papers
- PoseFormer · 2 of 16 papers
- STCFormer · 2 of 16 papers
- HSTFormer · 2 of 16 papers
- P-STMO · 2 of 16 papers
- WHAM · 2 of 16 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 60 papers; from OpenAlex and the papers' affiliations).
- Peking University · 5 papers
- Shanghai Jiao Tong University · 4 papers
- Tsinghua University · 4 papers
- Korea University · 3 papers
- Fudan University · 2 papers
- Beijing University of Posts and Telecommunications · 2 papers
- Xiamen University · 2 papers
- Meta · 2 papers
- Alibaba Group · 2 papers
- ShanghaiTech University · 2 papers
Use it
Download BibTeX of 39 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
- Monocular 2D-to-3D pose lifting (90 papers)
- Multi-person pose estimation (51 papers)
- Heatmap and coordinate pose encoding (50 papers)
- Temporal video pose estimation (76 papers)
Elsewhere in the topic
- Physics and sensor-based motion capture (125 papers)
- Learned camera pose estimation (96 papers)
- Category-level object pose estimation (92 papers)
- Depth-based hand pose regression (84 papers)
- Human shape and pose regression (80 papers)
- Instance-level 6D object pose (74 papers)
- Human mesh recovery from images (58 papers)
- Category-level pose estimation (54 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).