Research map / 3D vision / Human pose estimation
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use multi-hypothesis spatio-temporal modeling instead of a single-hypothesis Transformer (ablation) | MHFormer on Human3.6M Protocol 1, 27-frame input · Human3.6M | MPJPE (mm): 47.8 -> 45.9 | Helped |
| Add MHG to SHR-CHI (ablation) | SHR-CHI on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 47.2 -> 45.9 | Helped |
| Add the CHI module to MHG-SHR (ablation) | MHG-SHR on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 46.5 -> 45.9 | Helped |
| Add the SHR module to MHG-CHI (ablation) | MHG-CHI on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 46.8 -> 45.9 | Helped |
| Use multi-level features in MHG rather than parallel Transformer encoders (ablation) | MHFormer* on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 46.5 -> 45.9 | Helped |
| Increase MHG hypothesis count from 1 to 3 (ablation) | MHFormer on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 47.6 -> 45.9 | Helped |
| Increase MHG hypothesis count from 3 to 4 (ablation) | MHFormer on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 45.9 -> 46.9 | Hurt |
| Increase MHG layers from 2 to 4 (ablation) | MHFormer on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 46.4 -> 45.9 | Helped |
| Increase MHG layers from 4 to 5 (ablation) | MHFormer on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 45.9 -> 46.1 | Hurt |
| Set embedding dimension to 512 rather than 256 (ablation) | MHFormer on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 47.2 -> 45.9 | Helped |
| Increase embedding dimension from 512 to 768 (ablation) | MHFormer on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 45.9 -> 47.4 | Hurt |
| Use 3 SHR layers instead of 2 (ablation) | MHFormer on Human3.6M · Human3.6M, 27-frame input | MPJPE (mm): 45.9 -> 46.3 | Hurt |
Closest papers by meaning
- CHAMP: Conformalized 3D Human Multi-Hypothesis Pose Estimators (ICLR 2025 · similarity 0.90)
- ManiPose: Manifold-Constrained Multi-Hypothesis 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.89)
- SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · similarity 0.89)
- Multiple View Geometry Transformers for 3D Human Pose Estimation (CVPR 2024 · similarity 0.89)
- MV-SSM: Multi-View State Space Modeling for 3D Human Pose Estimation (CVPR 2025 · similarity 0.89)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.88)
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.