Research map / 3D vision / Human pose estimation
THUNDR: Transformer-Based 3D Human Reconstruction With Markers
Mihai Zanfir, Andrei Zanfir, Eduard Gabriel Bazavan, William T. Freeman, Rahul Sukthankar, Cristian Sminchisescu
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 intermediate marker representation instead of directly regressing GHUM parameters (ablation) | THUNDR · Human3.6M P2 (WS) | MPJPE-PA (mm): 63.5 -> 59.7 | Improved |
| add intermediate marker representation instead of directly regressing GHUM parameters (ablation) | THUNDR · Human3.6M P2 (WS) | MPJPE (mm): 95.4 -> 83.4 | Improved |
| use direct mesh Vd output instead of parametric mesh Vp for evaluation (ablation) | THUNDR · Human3.6M P2 (WS) | MPJPE-PA (mm): 61.8 -> 59.7 | Improved |
| use direct mesh Vd output instead of parametric mesh Vp for evaluation (ablation) | THUNDR · Human3.6M P2 (WS) | MPJPE (mm): 88.3 -> 83.4 | Improved |
| add full supervision (WS+FS) to weakly supervised training | THUNDR · Human3.6M P1 | MPJPE-PA (mm): 62.2 -> 39.8 | Improved |
| add full supervision (WS+FS) to weakly supervised training | THUNDR · 3DPW | MPJPE-PA (mm): 59.9 -> 51.5 | Improved |
| train marker poser with 5 mm Gaussian marker noise versus other levels (0,2,10,20,50 mm) (ablation) | marker poser · Human3.6M markers | reconstruction error | Improved |
| use trained marker poser instead of energy-minimization mesh fitting on Human3.6M markers | marker poser · Human3.6M mesh fitting | mean per-marker error (mm): 38.4 -> 44.3 | Mixed |
| replace HUND by THUNDR on official test | THUNDR · Human3.6M official test (900K samples) | MPJPE (mm): 66 -> 53 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SPIN — SPIN: Structure-Preserving Inner Offset Network for Scene Text Recognition (AAAI 2021) · better on 3 of 3 shared results
- I2L-MeshNet — I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image (ECCV 2020) · better on 3 of 3 shared results
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
- Neural Descent for Visual 3D Human Pose and Shape (CVPR 2021 · similarity 0.88)
- Cross-Attention of Disentangled Modalities for 3D Human Mesh Recovery with Transformers (ECCV 2022 · similarity 0.87)
- DiffSurf: A Transformer-based Diffusion Model for Generating and Reconstructing 3D Surfaces in Pose (ECCV 2024 · similarity 0.87)
- H4MER: Human 4D Modeling by Learning Neural Compositional Representation With Transformer (TPAMI 2023 · similarity 0.87)
- GeneMAN: Generalizable Single-Image 3D Human Reconstruction from Multi-Source Human Data (NeurIPS 2025 · similarity 0.86)
- Self-Supervised 3D Human Mesh Recovery from a Single Image with Uncertainty-Aware Learning (AAAI 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.
