Research map / 3D vision / 3D generation
AlignSDF: Pose-Aligned Signed Distance Fields for Hand-Object Reconstruction
Zerui Chen, Yana Hasson, Cordelia Schmid, Ivan Laptev
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 MANO-based pose normalization of query points for the hand SDF (model c vs a) (ablation) | hand SDF decoder · ObMan hand-only (87K train) | H_se: 0.128 -> 0.124 | Improved |
| add MANO-based pose normalization of query points for the hand SDF (model c vs a) (ablation) | hand SDF decoder · ObMan hand-only (30K train) | H_se: 0.183 -> 0.168 | Improved |
| add MANO network jointly trained with the SDF decoder (b vs a) (ablation) | hand SDF decoder · ObMan hand-only (87K train) | H_se: 0.128 -> 0.126 | Improved |
| add pose normalization on top of joint MANO training (c vs b) (ablation) | hand SDF decoder · ObMan hand-only (87K train) | H_je: 1.18 -> 1.20 | Hurt |
| use ground-truth hand pose instead of estimated pose (ablation) | AlignSDF hand · ObMan hand-only (87K train) | H_se: 0.124 -> 0.101 | Improved |
| add hand pose alignment to hand-object model (e vs d) (ablation) | hand-object SDF · ObMan hand-object (87K train) | H_se: 0.140 -> 0.131 | Improved |
| add hand pose alignment to hand-object model (e vs d) (ablation) | hand-object SDF · ObMan hand-object (87K train) | O_se: 4.09 -> 4.14 | Hurt |
| add object translation alignment to hand-object model (f vs d) (ablation) | hand-object SDF · ObMan hand-object (87K train) | O_se: 4.09 -> 3.36 | Improved |
| add object translation alignment alone (f vs d) (ablation) | hand-object SDF · ObMan hand-object (87K train) | H_se: 0.140 -> 0.148 | Hurt |
| combine hand and object alignment (g vs e) (ablation) | hand-object SDF · ObMan hand-object (87K train) | H_se: 0.131 -> 0.136 | Hurt |
| use ground-truth hand pose and object translation instead of estimates (g* vs g) (ablation) | AlignSDF · ObMan hand-object (87K train) | O_se: 3.38 -> 2.11 | Improved |
| replace Karunratanakul et al. 2De with AlignSDF | hand-object reconstruction · ObMan | H_se: 0.237 -> 0.136 | Improved |
Closest papers by meaning
- gSDF: Geometry-Driven Signed Distance Functions for 3D Hand-Object Reconstruction (CVPR 2023 · similarity 0.93)
- HOISDF: Constraining 3D Hand-Object Pose Estimation with Global Signed Distance Fields (CVPR 2024 · similarity 0.90)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.89)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 2024 · similarity 0.88)
- Hand-held Object Reconstruction from RGB Video with Dynamic Interaction (CVPR 2025 · similarity 0.88)
- HandBooster: Boosting 3D Hand-Mesh Reconstruction by Conditional Synthesis and Sampling of Hand-Object Interactions (CVPR 2024 · 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.
