Research map / 3D vision / Human pose estimation
Hallucinating Pose-Compatible Scenes
Tim Brooks, Alexei A. Efros
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 |
|---|---|---|---|
| remove style mixing (ablation) | pose-conditioned StyleGAN2 · Humans in Context | FID: 16.6 -> 11.6 | Improved |
| remove style mixing (ablation) | pose-conditioned StyleGAN2 · Humans in Context | PCKh: 32.4 -> 36.4 | Improved |
| add keypoint heatmap conditioning (ablation) | StyleGAN2 without style mixing · Humans in Context | PCKh: 36.4 -> 79.8 | Improved |
| add keypoint heatmap conditioning (ablation) | StyleGAN2 without style mixing · Humans in Context | FID: 11.6 -> 12.2 | Hurt |
| add augmentation and mismatch discrimination (ablation) | StyleGAN2 + heatmaps · Humans in Context | PCKh: 79.8 -> 80.7 | Improved |
| add augmentation and mismatch discrimination (ablation) | StyleGAN2 + heatmaps · Humans in Context | FID: 12.2 -> 12.1 | No clear effect |
| scale up model (2x channel width, batch 40 to 120) (ablation) | StyleGAN2 + heatmaps + augmentation · Humans in Context | FID: 12.1 -> 5.9 | Improved |
| scale up model (2x channel width, batch 40 to 120) (ablation) | StyleGAN2 + heatmaps + augmentation · Humans in Context | PCKh: 80.7 -> 84.2 | Improved |
| add pose latent conditioning to heatmap-only model (ablation) | heatmap-conditioned GAN · Humans in Context | FID: 15.1 -> 12.2 | Improved |
| add keypoint heatmaps to latent-only model (ablation) | latent-conditioned GAN · Humans in Context | PCKh: 36.4 -> 79.8 | Improved |
| use dual pose conditioning instead of heatmap only (ablation) | heatmap-conditioned GAN · Humans in Context | PCKh: 79.7 -> 79.8 | No clear effect |
| conditional truncation toward per-pose latent mean instead of global mean | Ours · qualitative | image quality | Improved |
Closest papers by meaning
- Putting People in Their Place: Affordance-Aware Human Insertion Into Scenes (CVPR 2023 · similarity 0.90)
- Scene-Aware Generative Network for Human Motion Synthesis (CVPR 2021 · similarity 0.89)
- Analyzing and Mitigating Object Hallucination: A Training Bias Perspective (AAAI 2026 · similarity 0.87)
- Can Generative Video Models Help Pose Estimation? (CVPR 2025 · similarity 0.86)
- Free-viewpoint Human Animation with Pose-correlated Reference Selection (CVPR 2025 · similarity 0.86)
- CapHuman: Capture Your Moments in Parallel Universes (CVPR 2024 · similarity 0.86)
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.
