Research map / 3D vision / Human pose estimation
AssemblyHands: Towards Egocentric Activity Understanding via 3D Hand Pose Estimation
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 |
|---|---|---|---|
| add iterative refinement that updates hand crops and volume root from the previous prediction (ablation) | MVExoNet annotation pipeline · Assembly101 annotation test | MPJPE (mm): 5.42~mm to 4.20~mm (22.5% reduction) | Helped |
| add iterative refinement for two rounds (ablation) | MVExoNet annotation pipeline · Desktop Activities | MPJPE (mm): 21.20~mm ... 13.38~mm (a 36.9% error reduction) | Helped |
| use multi-view volumetric feature fusion instead of standard 2D keypoint triangulation | MVExoNet annotation pipeline · Assembly101 annotation test | MPJPE (mm): 7.97 -> 4.20 | Helped |
| use exocentric multi-view annotation instead of egocentric-only tracking | 3D hand pose annotation · Assembly101 annotation test | MPJPE (mm): 27.55mm -> 4.20 | Helped |
| use 2D + Triangulation (OpenPose) | hand pose annotation · Assembly101 annotation test | PCK-AUC: 48.1 vs 83.4 | Hurt |
| train the egocentric estimator on automatic annotations alone rather than manual annotations (ablation) | SVEgoNet · Eval-M | MPJPE: 3% relative increase in error | Hurt |
| combine manual and automatic training annotations (ablation) | SVEgoNet · egocentric hand pose evaluation | MPJPE | Helped |
| replace UmeTrack poses with SVEgoNet poses | MS-G3D verb classifier · AssemblyHands-A | verb accuracy: 50.3 -> 54.7 | Helped |
| add object class labels to pose features (ablation) | pose-based verb classifier · AssemblyHands-A | verb accuracy: 54.7% -> 56.1% | Helped |
| replace UmeTrack poses with SVEgoNet poses | MS-G3D verb classifier · Assembly101, six verb classes | per-verb accuracy: pick up +2.1%; put down +6.2%; position +13.1%; screw +1.8%; unscrew +4.1%; remove -1.8% | Mixed |
Closest papers by meaning
- HOT3D: Hand and Object Tracking in 3D from Egocentric Multi-View Videos (CVPR 2025 · similarity 0.90)
- Single-to-Dual-View Adaptation for Egocentric 3D Hand Pose Estimation (CVPR 2024 · similarity 0.88)
- GigaHands: A Massive Annotated Dataset of Bimanual Hand Activities (CVPR 2025 · similarity 0.88)
- HandEdit: A Unified Benchmark for Egocentric Human-to-Robot Dexterous Hand Image Editing (NeurIPS 2026 · similarity 0.88)
- Modeling Fine-Grained Hand-Object Dynamics for Egocentric Video Representation Learning (ICLR 2025 · similarity 0.88)
- EgoPressure: A Dataset for Hand Pressure and Pose Estimation in Egocentric Vision (CVPR 2025 · 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.