Research map / 3D vision / 3D generation
Decoupled Generative Modeling for Human-Object Interaction Synthesis
Hwanhee Jung, Seunggwan Lee, Jeongyoon Yoon, SeungHyeon Kim, Giljoo Nam, Qixing Huang, Sangpil Kim
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 adversarial training on distal joints (ablation) | DecHOI baseline · FullBodyManipulation | hand penetration P_hand: 0.58 -> 0.53 | Improved |
| add adversarial training on distal joints (ablation) | DecHOI baseline · FullBodyManipulation | R-precision: 0.67 -> 0.64 | Hurt |
| add adversarial training on distal joints (ablation) | DecHOI baseline · FullBodyManipulation | T_e (cm): 7.92 -> 7.78 | Improved |
| add cross-attention text conditioning (ablation) | DecHOI baseline · FullBodyManipulation | R-precision: 0.67 -> 0.70 | Improved |
| add cross-attention text conditioning (ablation) | DecHOI baseline · FullBodyManipulation | contact F1: 0.65 -> 0.54 | Hurt |
| combine adversarial training and cross-attention (ablation) | DecHOI baseline · FullBodyManipulation | R-precision: 0.67 -> 0.72 | Improved |
| combine adversarial training and cross-attention (ablation) | DecHOI baseline · FullBodyManipulation | T_e (cm): 7.92 -> 6.91 | Improved |
| combine adversarial training and cross-attention (ablation) | DecHOI baseline · FullBodyManipulation | hand penetration P_hand: 0.58 -> 0.53 | Improved |
| decoupled trajectory/action generation with adversarial training vs single-network CHOIS | CHOIS · FullBodyManipulation | FID: 1.58 -> 0.33 | Improved |
| decoupled modeling vs CHOIS on unseen objects | CHOIS · 3D-FUTURE (unseen objects) | FID: 2.04 -> 1.01 | Improved |
| decoupled modeling vs CHOIS in dynamic long-sequence planning | CHOIS · DynaPlan (long-term dynamic) | human-scene penetration P_h->s: 0.83 -> 0.65 | Improved |
| decoupled modeling vs CHOIS in dynamic long-sequence planning | CHOIS · DynaPlan (long-term dynamic) | foot sliding FS: 0.43 -> 0.45 | Hurt |
Closest papers by meaning
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.90)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.90)
- NIFTY: Neural Object Interaction Fields for Guided Human Motion Synthesis (CVPR 2024 · similarity 0.90)
- HOI-Dyn: Learning Interaction Dynamics for Human-Object Motion Diffusion (NeurIPS 2025 · similarity 0.89)
- ChainHOI: Joint-based Kinematic Chain Modeling for Human-Object Interaction Generation (CVPR 2025 · similarity 0.89)
- InterControl: Zero-shot Human Interaction Generation by Controlling Every Joint (NeurIPS 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.
