Research map / 3D vision / 3D generation
ReGenHOI: Unifying Reconstruction and Generation for 3D Human-Object Interaction Understanding
Miao Xu, Xiangyu Zhu, Zidu Wang, Xusheng Liang, Bao Li, Jinlin Wu, Zelin Zang, Zhen Lei
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 reasoning trace mechanism (RTM) to contact prediction (ablation) | ReGenHOI · DAMON human contact | F1: 77.0 -> 78.4 | Improved |
| add reasoning trace mechanism (RTM) (ablation) | ReGenHOI · PICO reconstruction | PA-CD human (cm): 6.05 -> 5.42 | Improved |
| add gravity-based diffusion bridge (GBDB) refinement (ablation) | ReGenHOI · PICO reconstruction | Interaction volume IV (cm3): 1.20 -> 0.87 | Improved |
| add GBDB refinement (ablation) | ReGenHOI · PICO reconstruction | Penetration depth PD (mm): 2.45 -> 2.08 | Improved |
| add GBDB refinement (ablation) | ReGenHOI · PICO reconstruction | PA-CD object (cm): 12.84 -> 12.68 | Mixed |
| train reconstruction jointly with generation in shared latent space (ablation) | ReGenHOI · PICO reconstruction | PA-CD human (cm): 5.76 -> 5.42 | Improved |
| train reconstruction jointly with generation (ablation) | ReGenHOI · PICO reconstruction | PA-CD object (cm): 12.15 -> 12.68 | Hurt |
| train generation jointly with reconstruction (ablation) | ReGenHOI · FullBodyManipulation | Frechet Score FS: 0.49 -> 0.44 | Improved |
| train generation jointly with reconstruction (ablation) | ReGenHOI · FullBodyManipulation | FID: 1.03 -> 1.02 | No clear effect |
| train generation jointly with reconstruction (ablation) | ReGenHOI · BEHAVE | FID: 1.42 -> 1.38 | Improved |
| reason directly on 3D point clouds instead of 2D lifting | ReGenHOI · DAMON human contact | F1: 75.6 -> 78.4 | Improved |
| ReGenHOI vs InteractVLM per object category | ReGenHOI · DAMON semantic contact | F1: Food 66.4 -> 66.0; Sports 77.9 -> 78.4 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- InteractVLM — InteractVLM: 3D Interaction Reasoning from 2D Foundational Models (CVPR 2025) · better on 8 of 9 shared results
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
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.89)
- InteractAnything: Zero-shot Human Object Interaction Synthesis via LLM Feedback and Object Affordance Parsing (CVPR 2025 · similarity 0.88)
- Guiding Human-Object Interactions with Rich Geometry and Relations (CVPR 2025 · similarity 0.88)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.88)
- End-to-End HOI Reconstruction Transformer with Graph-based Encoding (CVPR 2025 · similarity 0.88)
- Reconstructing In-the-Wild Open-Vocabulary Human-Object Interactions (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.
