Research map / 3D vision / Human pose estimation
CrowdPose: Efficient Crowded Scenes Pose Estimation and a New Benchmark
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 joint-candidate SPPE and global association instead of the AlphaPose baseline | ResNet-101 pose estimation on CrowdPose · CrowdPose test set | mAP @0.5:0.95: 61.0 -> 66.0 | Helped |
| use joint-candidate SPPE and global association instead of the AlphaPose baseline | ResNet-101 pose estimation on CrowdPose · CrowdPose test set | mAR @0.5:0.95: 67.6 -> 72.7 | Helped |
| use joint-candidate SPPE and global association instead of the strongest listed baseline | ResNet-101 pose estimation · MSCOCO test-dev | mAP @0.5:0.95: 70.1 -> 70.9 | Helped |
| add joint-candidate loss (ablation) | CrowdPose · CrowdPose test set | mAP: 61.7 -> 66.0 | Helped |
| add joint-candidate loss (ablation) | CrowdPose · CrowdPose test set | mAR: 68.7 -> 72.7 | Helped |
| replace conventional Greedy NMS with global association (ablation) | CrowdPose · CrowdPose test set | mAP: 49.1 -> 66.0 | Helped |
| replace Parametric Pose-NMS with global association (ablation) | CrowdPose · CrowdPose test set | mAP: 64.2 -> 66.0 | Helped |
| replace conventional Greedy NMS with global association (ablation) | CrowdPose · CrowdPose test set | mAR: 63.1 -> 72.7 | Helped |
| replace Parametric Pose-NMS with global association (ablation) | CrowdPose · CrowdPose test set | mAR: 71.1 -> 72.7 | Helped |
Closest papers by meaning
- RSGNet: Relation based Skeleton Graph Network for Crowded Scenes Pose Estimation (AAAI 2021 · similarity 0.89)
- Robust Pose Estimation in Crowded Scenes with Direct Pose-Level Inference (NeurIPS 2021 · similarity 0.88)
- PoseTrack: Joint Multi-Person Pose Estimation and Tracking (CVPR 2017 · similarity 0.88)
- Crowd3D: Towards Hundreds of People Reconstruction From a Single Image (CVPR 2023 · similarity 0.87)
- ArtTrack: Articulated Multi-Person Tracking in the Wild (CVPR 2017 · similarity 0.87)
- DiffusionRegPose: Enhancing Multi-Person Pose Estimation using a Diffusion-Based End-to-End Regression Approach (CVPR 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.