Research map / 3D vision / 3D generation
Diffusion-Guided Reconstruction of Everyday Hand-Object Interaction Clips
Yufei Ye, Poorvi Hebbar, Abhinav Gupta, Shubham Tulsiani
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 the learned diffusion prior (category and hand conditioned) via SDS (ablation) | DiffHOI · HOI4D (mean) | Chamfer distance: 2.7 -> 0.8 | Improved |
| add category conditioning to the diffusion prior (ablation) | DiffHOI · HOI4D (mean) | Chamfer distance: 2.8 -> 0.8 | Improved |
| use hand-only prior instead of no prior (ablation) | DiffHOI · HOI4D (mean) | F@5: 0.47 -> 0.39 | Hurt |
| add hand conditioning to the prior for hand-object alignment (ablation) | DiffHOI · HOI4D (mean) | CD_h: 85.2 -> 48.7 | Improved |
| distill surface normals in SDS (ablation) | DiffHOI · HOI4D (mean) | Chamfer distance: 4.3 -> 0.8 | Improved |
| distill masks in SDS (ablation) | DiffHOI · HOI4D (mean) | Chamfer distance: 1.3 -> 0.8 | Improved |
| distill depth in SDS (ablation) | DiffHOI · HOI4D (mean) | F@5: 0.66 -> 0.62 | Mixed |
| soft blending instead of hard blending with ordinal depth loss (ablation) | DiffHOI · HOI4D (mean) | CD_h: 87.8 -> 48.7 | Improved |
| optimize per-frame object poses (ablation) | DiffHOI · HOI4D (mean) | CD_h: 67.4 -> 48.7 | Improved |
| doubling hand pose prediction error (ablation) | DiffHOI · HOI4D (mean) | Chamfer distance: 0.77 -> 1.01 | Hurt |
| diffusion-guided multi-frame optimization instead of single-frame feed-forward | iHOI baseline · HOI4D (mean) | Chamfer distance: 2.7 -> 0.8 | Improved |
Closest papers by meaning
- The Surprising Effectiveness of Video Diffusion Models for Hand Motion Reconstruction (NeurIPS 2026 · similarity 0.91)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.90)
- Re-HOLD: Video Hand Object Interaction Reenactment via adaptive Layout-instructed Diffusion Model (CVPR 2025 · similarity 0.90)
- HOIDiffusion: Generating Realistic 3D Hand-Object Interaction Data (CVPR 2024 · similarity 0.90)
- BIGS: Bimanual Category-agnostic Interaction Reconstruction from Monocular Videos via 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
- HandBooster: Boosting 3D Hand-Mesh Reconstruction by Conditional Synthesis and Sampling of Hand-Object Interactions (CVPR 2024 · 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.
