Research map / 3D vision / Human pose estimation
Mesh Graphormer
Kevin Lin, Lijuan Wang, Zicheng Liu
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 image grid features (ablation) | Transformer · Human3.6M | PA-MPJPE: 36.7 -> 35.9 | Improved |
| add graph convolution to Encoder1 (ablation) | Graphormer baseline · Human3.6M | PA-MPJPE: 35.9 -> 36.7 | No clear effect |
| add graph convolution to Encoder2 (ablation) | Graphormer baseline · Human3.6M | PA-MPJPE: 35.9 -> 36.2 | Hurt |
| add graph convolution to Encoder3 (ablation) | Graphormer baseline · Human3.6M | PA-MPJPE: 35.9 -> 35.1 | Improved |
| add graph convolution to all three encoders (ablation) | Graphormer baseline · Human3.6M | PA-MPJPE: 35.9 -> 36.0 | Hurt |
| place graph convolution after MHSA instead of in parallel (ablation) | Graphormer block · Human3.6M | PA-MPJPE: 36.4 -> 35.1 | Improved |
| place graph convolution after MHSA instead of before (ablation) | Graphormer block · Human3.6M | PA-MPJPE: 35.6 -> 35.1 | Improved |
| replace basic graph convolution with Graph Residual Block (ablation) | Graphormer with grid features · Human3.6M | PA-MPJPE: 35.1 -> 34.5 | Improved |
| add graph convolution without grid features (ablation) | Transformer · Human3.6M | PA-MPJPE: 36.7 -> 36.7 | No clear effect |
| add graph convolution and residual block together with grid features (ablation) | Transformer + Grid Features · Human3.6M | PA-MPJPE: 36.7 -> 34.5 | Improved |
| replace graph residual block with larger MLP (ablation) | Graphormer · Human3.6M | PA-MPJPE: 35.9 -> 36.0 | No clear effect |
| use Mesh Graphormer instead of METRO | METRO · 3DPW | PA-MPJPE: 47.9 -> 45.6 | 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.
- VIBE — VIBE: Video Inference for Human Body Pose and Shape Estimation (CVPR 2020) · better on 5 of 5 shared results
- I2LMeshNet — I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image (ECCV 2020) · better on 2 of 2 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
- GraFormer: Graph-Oriented Transformer for 3D Pose Estimation (CVPR 2022 · similarity 0.90)
- End-to-End Human Pose and Mesh Reconstruction with Transformers (CVPR 2021 · similarity 0.88)
- Pose2Mesh: Graph Convolutional Network for 3D Human Pose and Mesh Recovery from a 2D Human Pose (ECCV 2020 · similarity 0.87)
- End-to-End HOI Reconstruction Transformer with Graph-based Encoding (CVPR 2025 · similarity 0.86)
- Multiple View Geometry Transformers for 3D Human Pose Estimation (CVPR 2024 · similarity 0.85)
- MeshGPT: Generating Triangle Meshes with Decoder-Only Transformers (CVPR 2024 · 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.
