Research map / 3D vision / Face and shape
Coarse-to-Fine Implicit Representation Learning for 3D Hand-Object Reconstruction from a Single RGB-D Image
Xingyu Liu, Pengfei Ren, Jingyu Wang, Qi Qi, Haifeng Sun, Zirui Zhuang, Jianxin Liao
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 |
|---|---|---|---|
| replace pixel-aligned features with 3D point-aligned features (3D PIFu) (ablation) | RGB-D SDF reconstruction · DexYCB | Hand Chamfer distance CDh (cm2): 0.299 -> 0.280 | Improved |
| add point cloud feature refinement via coarse-level SDF (ablation) | 3D PIFu model · DexYCB | Object Chamfer distance CDo (cm2): 1.56 -> 1.26 | Improved |
| add surface-aware efficient reconstruction (sparse SDF queries) (ablation) | full model · DexYCB | Reconstruction time (s): 3.44 -> 1.26 s; 2.1M -> 89.8K sampling points | Improved |
| add SER to full model (ablation) | full model · DexYCB | Hand Chamfer distance CDh (cm2): 0.272 -> 0.267 (Table 1 ID 3 vs 4); identical 0.267 with and without sparse sampling | No clear effect |
| use heatmap only for sparse queries instead of heatmap plus point cloud (ablation) | SER · DexYCB | Hand Chamfer distance CDh (cm2): 0.267 -> 0.279 | Hurt |
| add RGB-D input instead of RGB only with pixel-aligned features (ablation) | SDF baseline · DexYCB | Object Chamfer distance CDo (cm2): 2.04 -> 1.50 | Improved |
| use proposed method instead of adding SA-Gate RGB-D fusion to gSDF | RGB-D hand-object reconstruction · DexYCB | Object Chamfer distance CDo (cm2): 1.28 -> 1.24 | Improved |
| use coarse-to-fine RGB-D SDF instead of gSDF | ObMan · ObMan | Object Chamfer distance CDo (cm2): 3.14 -> 0.51 | 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.
- gSDF — GSDF: 3DGS Meets SDF for Improved Neural Rendering and Reconstruction (NeurIPS 2024) · better on 8 of 8 shared results
- DDF-HO — DDF-HO: Hand-Held Object Reconstruction via Conditional Directed Distance Field (NeurIPS 2023) · better on 1 of 1 shared result
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
- gSDF: Geometry-Driven Signed Distance Functions for 3D Hand-Object Reconstruction (CVPR 2023 · similarity 0.92)
- AlignSDF: Pose-Aligned Signed Distance Fields for Hand-Object Reconstruction (ECCV 2022 · similarity 0.91)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 2024 · similarity 0.89)
- SRHand: Super-Resolving Hand Images and 3D Shapes via View/Pose-aware Neural Image Representations and Explicit Meshes (NeurIPS 2025 · similarity 0.88)
- HOISDF: Constraining 3D Hand-Object Pose Estimation with Global Signed Distance Fields (CVPR 2024 · similarity 0.88)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.87)
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.
