Research map / 3D vision / Human pose estimation
Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation
Read the paperCheck 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 |
|---|---|---|---|
| Replace PAFs with body-part heatmaps | 3-stage CMU-Net · MS-COCO 2017 validation | AP: 56.0 -> 56.5 | Helped |
| Add focal L2 loss (ablation) | 3-stage CMU-Net with body-part heatmaps · MS-COCO 2017 validation | AP: 56.5 -> 60.7 | Helped |
| Apply focal L2 to body-part heatmaps as well as keypoint heatmaps (ablation) | 3-stage IMHN · MS-COCO 2017 validation | AP: 64.2 -> 65.8 | Helped |
| Add spatial attention (ablation) | 3-stage IMHN · MS-COCO 2017 validation | AP: 64.6 -> 65.8 | Helped |
| Add channel attention (ablation) | 3-stage IMHN · MS-COCO 2017 validation | AP: 65.4 -> 65.8 | Helped |
| Add redundant body-part connections to the skeleton (ablation) | 3-stage IMHN · MS-COCO 2017 validation | AP: 65.1 -> 65.8 | Helped |
| Replace the 3-stage CMU-Net with 3-stage IMHN (ablation) | Proposed bottom-up pose estimator · MS-COCO 2017 validation | AP: 60.7 -> 65.8 | Helped |
| Increase from 3-stage to 4-stage IMHN (ablation) | Proposed bottom-up pose estimator · MS-COCO 2017 test-dev | AP: 65.2 -> 66.2 | Helped |
| Increase input size and use the plus model configuration (ablation) | 4-stage IMHN · MS-COCO 2017 test-dev | AP: 66.2 -> 68.1 | Helped |
| Train with MS-COCO validation data (ablation) | 4-stage IMHN plus · MS-COCO 2017 validation | AP: 69.1 -> 74.1 | Helped |
| Use one-scale rather than multi-scale inference (ablation) | 4-stage IMHN plus · MS-COCO 2017 validation | AP: 69.1 -> 65.9 | Hurt |
Closest papers by meaning
- Towards Accurate Multi-Person Pose Estimation in the Wild (CVPR 2017 · similarity 0.92)
- DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation (CVPR 2016 · similarity 0.91)
- SIMPLE: SIngle-network with Mimicking and Point Learning for Bottom-up Human Pose Estimation (AAAI 2021 · similarity 0.91)
- AdaptivePose: Human Parts as Adaptive Points (AAAI 2022 · similarity 0.91)
- Monocular 3D Multi-Person Pose Estimation by Integrating Top-Down and Bottom-Up Networks (CVPR 2021 · similarity 0.90)
- Cascaded Pyramid Network for Multi-Person Pose Estimation (CVPR 2018 · similarity 0.90)
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.