Research map / 3D vision / Human pose estimation
InterHand2.6M: A Dataset and Baseline for 3D Interacting Hand Pose Estimation from a Single RGB Image
Gyeongsik Moon, Shoou-I Yu, He Wen, Takaaki Shiratori, Kyoung Mu Lee
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 interacting hand data to single hand training data (ablation) | InterNet · InterHand2.6M | interacting hand MPJPE: 51.19 -> 16.02 | Improved |
| add interacting hand data to single hand training data (ablation) | InterNet · InterHand2.6M | single hand MPJPE: 13.08 -> 12.16 | Improved |
| add single hand data to interacting hand data (ablation) | InterNet · InterHand2.6M | interacting hand MPJPE: 16.86 -> 16.02 | Improved |
| add machine annotations to human annotations (Train H+M vs H) (ablation) | InterNet · InterHand2.6M Test (H+M) | interacting hand MPJPE: 17.16 -> 16.02 | Improved |
| add machine annotations to human annotations (Train H+M vs H) (ablation) | InterNet · InterHand2.6M Test (H+M) | single hand MPJPE: 12.58 -> 12.16 | Improved |
| use machine-only instead of human-only annotation (ablation) | InterNet · InterHand2.6M Test (H) | interacting hand MPJPE: 13.05 -> 14.26 | Hurt |
| use right hand-relative left hand depth when right hand visible instead of always z^L (ablation) | InterNet · InterHand2.6M Test (H+M) | MRRPE: 92.14 -> 32.57 | Improved |
| train with all views instead of 4 views (ablation) | InterNet · InterHand2.6M Test (H+M) | MPJPE: 26.80 -> 14.22 | Improved |
| train with all views instead of 4 views (ablation) | InterNet · InterHand2.6M Test (H+M) | MRRPE: 61.90 -> 32.57 | Improved |
| increase number of input views for machine annotator triangulation (ablation) | machine annotator · Test (H) | 3D distance error: 2.78 mm with all 90 views | Improved |
| use InterNet 3D heatmaps without GT scale or handedness vs Spurr et al. without GT | InterNet · STB | EPE: 9.49 -> 7.95 | Improved |
| use InterNet vs Spurr et al. with GT on RHP | InterNet · RHP | EPE: 19.73 -> 20.89 | Hurt |
Closest papers by meaning
- First-Person Hand Action Benchmark With RGB-D Videos and 3D Hand Pose Annotations (CVPR 2018 · similarity 0.92)
- End-to-End Detection and Pose Estimation of Two Interacting Hands (ICCV 2021 · similarity 0.92)
- FreiHAND: A Dataset for Markerless Capture of Hand Pose and Shape From Single RGB Images (ICCV 2019 · similarity 0.92)
- SeqHAND: RGB-Sequence-Based 3D Hand Pose and Shape Estimation (ECCV 2020 · similarity 0.91)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.89)
- BIGS: Bimanual Category-agnostic Interaction Reconstruction from Monocular Videos via 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
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.
