Research map / 3D vision / Human pose estimation
Rethinking Keypoint Representations: Modeling Keypoints and Poses as Objects for Multi-Person Human Pose Estimation
William McNally, Kanav Vats, Alexander Wong, John McPhee
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 keypoint object fusion to pose objects (ablation) | KAPAO-S · COCO val2017 (no TTA) | AP: +1.7 ms / +2.4 AP | Improved |
| add keypoint object fusion to pose objects (ablation) | KAPAO-M · COCO val2017 (no TTA) | AP: +1.2 ms / +1.1 AP | Improved |
| add keypoint object fusion to pose objects (ablation) | KAPAO-L · COCO val2017 (no TTA) | AP: +1.7 ms / +1.2 AP | Improved |
| add keypoint object fusion to pose objects (ablation) | KAPAO-S · CrowdPose test (no TTA) | AP: +3.3 ms / +2.9 AP | Improved |
| add keypoint object fusion to pose objects (ablation) | KAPAO-L · CrowdPose test (with TTA) | AP: +4.5 ms / +1.0 AP | Improved |
| use keypoint box size above 5% of max(w,h) (ablation) | KAPAO-S · COCO train2017 50 epochs | AP | Hurt |
| use keypoint box size below 2.5% of max(w,h) (ablation) | KAPAO-S · COCO train2017 50 epochs | AP | Hurt |
| lower keypoint confidence threshold tau_ck to 0.01 (ablation) | KAPAO-L · COCO val2017 (no TTA) | AP: 70.4 -> 70.1 | Hurt |
| reduce input size from 1280 to 1152 (ablation) | KAPAO-S/M/L · COCO val2017 (no TTA) | AP | No clear effect |
| reduce input size to 1024 (ablation) | KAPAO-M/L · COCO val2017 (no TTA) | latency: reduced the latency by ~30% | Mixed |
| add multi-scale and flip TTA | KAPAO-L · COCO val2017 | AP: 70.6 -> 71.6 | Improved |
| replace HigherHRNet-W48 + SWAHR by KAPAO-L at test | KAPAO-L · COCO test-dev | AP: 1.7 AP lower, 7.4x faster | Mixed |
Closest papers by meaning
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.89)
- Keypoint Communities (ICCV 2021 · similarity 0.89)
- Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation (AAAI 2020 · similarity 0.88)
- Distribution-Aware Coordinate Representation for Human Pose Estimation (CVPR 2020 · similarity 0.87)
- Bottom-Up Human Pose Estimation via Disentangled Keypoint Regression (CVPR 2021 · similarity 0.87)
- QueryPose: Sparse Multi-Person Pose Regression via Spatial-Aware Part-Level Query (NeurIPS 2022 · 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.
