Research map / 3D vision / Human pose estimation
PifPaf: Composite Fields for Human Pose Estimation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PifPaf instead of Mask R-CNN | COCO low-resolution pose estimation · COCO low resolution (long side 321 px) | AP: 41.6 -> 50.0 | Helped |
| Use PifPaf instead of OpenPose | COCO low-resolution pose estimation · COCO low resolution (long side 321 px) | AP: 37.6 -> 50.0 | Helped |
| Use PifPaf instead of PersonLab single-scale | COCO pose estimation · COCO 2017 test-dev | AP: 66.5 -> 66.7 | Helped |
| Use PifPaf instead of Mask R-CNN | Market-1501 pose extraction · Market-1501 train set (202 random images) | accuracy: 43% -> 96% | Helped |
| Replace vanilla L1 with Laplace regression loss (ablation) | PifPaf · COCO validation (20 epochs, ResNet50) | AP: 41.7 -> 45.1 | Helped |
| Use SmoothL1 with r=0.2√(A_i)σ_k instead of vanilla L1 (ablation) | PifPaf · COCO validation (20 epochs, ResNet50) | AP: 41.7 -> 42.0 | Helped |
| Use SmoothL1 with r=0.5√(A_i)σ_k instead of vanilla L1 (ablation) | PifPaf · COCO validation (20 epochs, ResNet50) | AP: 41.7 -> 41.9 | No clear effect |
| Use SmoothL1 with r=1.0√(A_i)σ_k instead of vanilla L1 (ablation) | PifPaf · COCO validation (20 epochs, ResNet50) | AP: 41.7 -> 41.6 | No clear effect |
| Use the predicted spread b in the decoder (ablation) | PifPaf with Laplace loss · COCO validation (20 epochs, ResNet50) | AP: 45.1 -> 45.5 | Helped |
| Add the scale component σ to the PIF field (ablation) | PifPaf · COCO | AP: 64.5 -> 65.7 | Helped |
Closest papers by meaning
- Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields (CVPR 2017 · similarity 0.89)
- Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation (AAAI 2020 · similarity 0.86)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.84)
- Reconstructing People, Places, and Cameras (CVPR 2025 · similarity 0.83)
- AiOS: All-in-One-Stage Expressive Human Pose and Shape Estimation (CVPR 2024 · similarity 0.83)
- SHaRPose: Sparse High-Resolution Representation for Human Pose Estimation (AAAI 2024 · similarity 0.83)
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.