Research map / 3D vision / 3D generation
Contact-Aware Amodal Completion for Human-Object Interaction via Multi-Regional Inpainting
Seunggeun Chi, Enna Sachdeva, Pin-Hao Huang, Kwonjoon Lee
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 |
|---|---|---|---|
| use multi-regional inpainting with primary and secondary regions instead of a single contact-aware hull region (ablation) | contact-aware inpainting · BEHAVE | mIoU (%): 76.11% -> 77.64% | Improved |
| add contact points to the convex hull (ablation) | single-region hull inpainting · BEHAVE | mIoU (%): 75.24% -> 76.11% | Improved |
| use contact-aware convex hull region instead of whole human mask (ablation) | single-region inpainting · BEHAVE | mIoU (%): 69.98% -> 76.11% | Improved |
| restrict inpainting to hull region instead of naive outpainting of whole canvas (ablation) | SD-inpaint · BEHAVE | mIoU (%): 50.92% -> 76.11% | Improved |
| restrict inpainting to hull region instead of naive outpainting (ablation) | SD-inpaint · BEHAVE | CLIP score: 27.34 -> 26.63 | Hurt |
| use ground-truth contact instead of VLM-estimated contact (ablation) | Ours · BEHAVE | mIoU (%): 77.64% -> 80.15% | Improved |
| use VLM contact estimation instead of DECO (ablation) | Ours with Multi-HMR · BEHAVE | object mIoU (%): 75.02% -> 77.64% | Improved |
| add DECO contact to no-contact pipeline (ablation) | Ours with Multi-HMR · BEHAVE | object mIoU (%): 74.80% -> 75.02% | No clear effect |
| lower strength r from 0.5 to 0.1 for light occlusion (ablation) | Ours · BEHAVE (occ. 10-40%) | mIoU (%): 84.70% -> 85.44% | Improved |
| raise strength r from 0.5 to 0.9 for heavy occlusion (ablation) | Ours · BEHAVE (occ. 40-70%) | mIoU (%): 72.93% -> 73.94% | Improved |
| use r=0.5 instead of r=0.0 (add secondary region inpainting) (ablation) | Ours · BEHAVE (total) | mIoU (%): 76.11% -> 77.64% | Improved |
| set r=1.0 instead of 0.5 (ablation) | Ours · BEHAVE (total) | mIoU (%): 77.64% -> 69.98% | Hurt |
Closest papers by meaning
- Amodal Scene Analysis via Holistic Occlusion Relation Inference and Generative Mask Completion (AAAI 2024 · similarity 0.89)
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.88)
- Using Diffusion Priors for Video Amodal Segmentation (CVPR 2025 · similarity 0.88)
- Open-World Amodal Appearance Completion (CVPR 2025 · similarity 0.87)
- InteractVLM: 3D Interaction Reasoning from 2D Foundational Models (CVPR 2025 · similarity 0.87)
- In-Hand 3D Object Reconstruction from a Monocular RGB Video (AAAI 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.
