Research map / 3D vision / 3D generation
CHORUS : Learning Canonicalized 3D Human-Object Spatial Relations from Unbounded Synthesized Images
Sookwan Han, Hanbyul Joo
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 pose canonicalization before aggregating 2D cues (ablation) | CHORUS · COCO-EFT categories (vanilla) | mPAP: 17.52 -> 19.86 | Improved |
| add keypoint filtering of synthesized images (ablation) | CHORUS · COCO-EFT categories (vanilla) | mPAP: 18.31 -> 19.86 | Improved |
| use more synthesized images (10% to 100%) (ablation) | CHORUS · COCO-EFT categories (vanilla) | mPAP: 17.90 -> 19.86 | Improved |
| use 1% instead of full synthesized data (ablation) | CHORUS · COCO-EFT categories (vanilla) | mPAP: 19.86 -> 14.98 | Hurt |
| synthesize images with diffusion instead of internet image search (10k) (ablation) | CHORUS aggregation on motorcycle · motorcycle (vanilla) | mPAP: 55.97 -> 57.14 | Improved |
| synthesize images instead of image search at 10k (ablation) | data collection · motorcycle image retrieval | rejection rate (%): 73.15 -> 61.46 | Improved |
| synthesize images instead of image search at 10k (ablation) | data collection · motorcycle image retrieval | camera entropy (bits): 12.25 -> 14.63 | Improved |
| scale image search from 1k to 10k images (ablation) | Image Search baseline · motorcycle (vanilla) | mPAP: 53.64 -> 55.97 | Improved |
| scale synthetic data from 1k to 10k images (ablation) | CHORUS · motorcycle (strict, vanilla) | mPAP_strict: 55.11 -> 55.19 | No clear effect |
Closest papers by meaning
- InteractAnything: Zero-shot Human Object Interaction Synthesis via LLM Feedback and Object Affordance Parsing (CVPR 2025 · similarity 0.90)
- LEMON: Learning 3D Human-Object Interaction Relation from 2D Images (CVPR 2024 · similarity 0.89)
- BOP-ASK: Object-Interaction Reasoning for Vision-Language Models (CVPR 2026 · similarity 0.88)
- CG-HOI: Contact-Guided 3D Human-Object Interaction Generation (CVPR 2024 · similarity 0.88)
- Template Free Reconstruction of Human-object Interaction with Procedural Interaction Generation (CVPR 2024 · similarity 0.88)
- Can Transformers Capture Spatial Relations between Objects? (ICLR 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.
