Research map / 3D vision / 3D generation
ContactGen: Generative Contact Modeling for Grasp Generation
Shaowei Liu, Yang Zhou, Jimei Yang, Saurabh Gupta, Shenlong Wang
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 |
|---|---|---|---|
| represent contact with part and direction maps in addition to contact map (ContactGen vs TOCH) | contact-to-grasp recovery · GRAB (recovery from GT contact) | EPE (cm): 3.44 -> 1.49 | Improved |
| use sequential (hierarchical) CVAE instead of joint modeling of the three maps (ablation) | ContactGen CVAE · GRAB grasp synthesis | Simulation displacement: 3.29 -> 2.16 | Improved |
| use sequential CVAE instead of separate modeling of the three maps (ablation) | ContactGen CVAE · GRAB grasp synthesis | Contact ratio: 0.89 -> 0.96 | Improved |
| sequential decomposition versus separate modeling on diversity (ablation) | ContactGen CVAE · GRAB grasp synthesis | Cluster size: 4.83 -> 4.11 | Hurt |
| add part map and direction map to contact map (SDF hand) (ablation) | ContactGen contact solver · GRAB grasp synthesis | Penetration volume: 20.13 -> 2.72 | Improved |
| add direction map on top of contact and part maps (SDF hand) (ablation) | ContactGen contact solver · GRAB grasp synthesis | Simulation displacement: 6.81 -> 2.16 | Improved |
| add part and direction maps (MANO hand) (ablation) | contact solver with MANO · GRAB grasp synthesis | Penetration volume: 7.41 -> 2.36 | Improved |
| use piecewise SDF hand instead of MANO with full representation (ablation) | contact solver · GRAB grasp synthesis | Simulation displacement: 2.85 -> 2.16 | Improved |
| use ContactGen instead of HALO on HO3D | grasp synthesis · HO3D grasp synthesis (out-of-domain) | Penetration volume: 25.84 -> 9.96 | Improved |
| use ContactGen instead of GrabNet on GRAB | grasp synthesis · GRAB grasp synthesis | Simulation displacement: 1.72 -> 2.16 | Hurt |
| use ContactGen instead of HALO in human study (GRAB naturalness) | grasp synthesis · GRAB human study | Natural score: 2.59 -> 3.21 | Improved |
Later papers that compared with this method
6 later accepted papers report ContactGen as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- TOUCH: Text-guided Controllable Generation of Free-Form Hand-Object Interactions (ICLR 2026 · better on 12 of 12 shared results)
- Force-Aware 3D Contact Modeling for Stable Grasp Generation (AAAI 2026 · better on 6 of 6 shared results)
- VolCo: Volumetric Contact for High-Fidelity Human Grasp Generation (NeurIPS 2026 · better on 4 of 4 shared results)
- Task-Oriented Human Grasp Synthesis via Context- and Task-Aware Diffusers (ICCV 2025 · better on 2 of 2 shared results)
- Decomposed Vector-Quantized Variational Autoencoder for Human Grasp Generation (ECCV 2024 · better on 3 of 4 shared results)
- NL2Contact: Natural Language Guided 3D Hand-Object Contact Modeling with Diffusion Model (ECCV 2024 · better on 2 of 2 shared results)
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.
- TOCH — TOCH: Spatio-Temporal Object-to-Hand Correspondence for Motion Refinement (ECCV 2022) · better on 4 of 4 shared results
- ContactOpt — ContactOpt: Optimizing Contact To Improve Grasps (CVPR 2021) · 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
- Contact Map Transfer with Conditional Diffusion Model for Generalizable Dexterous Grasp Generation (NeurIPS 2025 · similarity 0.90)
- RAGG: Retrieval-Augmented Grasp Generation Model (AAAI 2025 · similarity 0.88)
- Text2HOI: Text-guided 3D Motion Generation for Hand-Object Interaction (CVPR 2024 · similarity 0.87)
- G-HOP: Generative Hand-Object Prior for Interaction Reconstruction and Grasp Synthesis (CVPR 2024 · similarity 0.86)
- VolCo: Volumetric Contact for High-Fidelity Human Grasp Generation (NeurIPS 2026 · similarity 0.86)
- LatentHOI: On the Generalizable Hand Object Motion Generation with Latent Hand Diffusion. (CVPR 2025 · similarity 0.86)
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.
