Research map / 3D vision / Human pose estimation
Deformable Mesh Transformer for 3D Human Mesh Recovery
Read the paperCheck 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 DeFormer with HRNet-w48 instead of FastMETRO | 3DPW · 3DPW | MPVE: 84.1 -> 82.6 | Helped |
| Use DeFormer with HRNet-w48 instead of FastMETRO | 3DPW · 3DPW | MPJPE: 73.5 -> 72.9 | Helped |
| Use DeFormer with HRNet-w48 instead of FastMETRO | 3DPW · 3DPW | PA-MPJPE: 44.6 -> 44.3 | Helped |
| Use DeFormer with HRNet-w48 instead of FastMETRO | Human3.6M P2 · Human3.6M P2 | MPJPE: 33.7 -> 31.6 | Helped |
| Use four HRNet feature-map scales rather than only W4 (ablation) | DeFormer with HRNet-w48-COCO · Human3.6M | MPJPE: 50.9 -> 44.8 | Helped |
| Use four HRNet feature-map scales rather than only W4 (ablation) | DeFormer with HRNet-w48-COCO · Human3.6M | PA-MPJPE: 36.5 -> 31.6 | Helped |
| Use four ResNet50 feature-map scales rather than only W4 (ablation) | DeFormer with ResNet50-COCO · Human3.6M | MPJPE: 50.5 -> 50.7 | Mixed |
| Replace MS deformable cross-attention with bilinear interpolation (ablation) | DeFormer · Human3.6M | MPJPE: 44.8 -> 48.3 | Hurt |
| Replace MS deformable cross-attention with standard transformer cross-attention (ablation) | DeFormer · Human3.6M | MPJPE: 44.8 -> 50.7 | Hurt |
| Use three MS deformable cross-attention layers instead of one (ablation) | DeFormer · Human3.6M | MPJPE: 47.7 -> 44.8 | Helped |
| Replace MS deformable cross-attention with bilinear interpolation (ablation) | DeFormer · Human3.6M | PA-MPJPE: 31.6 -> 34.3 | Hurt |
| Replace MS deformable cross-attention with standard transformer cross-attention (ablation) | DeFormer · Human3.6M | PA-MPJPE: 31.6 -> 37.1 | Hurt |
Closest papers by meaning
- PostureHMR: Posture Transformation for 3D Human Mesh Recovery (CVPR 2024 · similarity 0.86)
- GenHMR: Generative Human Mesh Recovery (AAAI 2025 · similarity 0.85)
- FactorizedHMR: A Hybrid Framework for Video Human Mesh Recovery (NeurIPS 2026 · similarity 0.85)
- MeshPose: Unifying DensePose and 3D Body Mesh Reconstruction (CVPR 2024 · similarity 0.84)
- MEGA: Masked Generative Autoencoder for Human Mesh Recovery (CVPR 2025 · similarity 0.84)
- ADHMR: Aligning Diffusion-based Human Mesh Recovery via Direct Preference Optimization (ICML 2025 · similarity 0.84)
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.