Research map / 3D vision / Human pose estimation
Generating 3D People in Scenes Without People
Yan Zhang, Mohamed Hassan, Heiko Neumann, Michael J. Black, Siyu Tang
Check 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 a scene-conditioned S1 CVAE instead of the modified baseline generator | 3D human-scene generation · PROX-E test | reconstruction error: 0.48 -> 0.25 | Improved |
| Train with human-scene interaction loss (ablation) | S1 generator · PROX-E val | reconstruction error: 0.22 -> 0.16 | Improved |
| Train with human-scene interaction loss (ablation) | S1 generator · PROX-E test | reconstruction error: 0.25 -> 0.24 | No clear effect |
| Train with human-scene interaction loss (ablation) | S1 generator · MP3D-R | contact score: 0.80 -> 0.65 | Hurt |
| Use two-stage factorized architecture instead of one-stage (ablation) | CVAE generator · PROX-E test | reconstruction error: 0.25 -> 0.70 | Hurt |
| Apply scene geometry-aware fitting (ablation) | S1 generator · PROX-E | non-collision score: 0.93 -> 0.94 | Improved |
| Apply scene geometry-aware fitting (ablation) | Baseline generator · MP3D-R | contact score: 0.78 -> 0.89 | Improved |
| Apply scene geometry-aware fitting (ablation) | S1 generator · PROX-E | user study score: 3.29 ± 1.36 -> 3.49 ± 1.26 | Improved |
| Use the learned scene-dependent pose prior for pose optimization | 3D pose estimation · PROX-Quantitative | V2V: 173.97 -> 171.75 | Improved |
| Use the learned scene-dependent pose prior for pose optimization | 3D pose estimation · PROX-Quantitative | PJE: 171.78 -> 174.10 | Hurt |
Closest papers by meaning
- Pose-Guided 3D Human Generation in Indoor Scene (AAAI 2023 · similarity 0.91)
- Populating 3D Scenes by Learning Human-Scene Interaction (CVPR 2021 · similarity 0.89)
- InteractAnything: Zero-shot Human Object Interaction Synthesis via LLM Feedback and Object Affordance Parsing (CVPR 2025 · similarity 0.89)
- GenZI: Zero-Shot 3D Human-Scene Interaction Generation (CVPR 2024 · similarity 0.89)
- En3D: An Enhanced Generative Model for Sculpting 3D Humans from 2D Synthetic Data (CVPR 2024 · similarity 0.88)
- Template Free Reconstruction of Human-object Interaction with Procedural Interaction Generation (CVPR 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.
