Research map / 3D vision / Human pose estimation
Simple Baselines for Human Pose Estimation and Tracking
Bin Xiao, Haiping Wu, Yichen Wei
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 |
|---|---|---|---|
| use three deconvolutional layers instead of two (ablation) | ResNet-50 deconv head · COCO val2017 | AP: 67.9 -> 70.4 | Improved |
| increase deconv kernel size from 2 to 4 (ablation) | ResNet-50 deconv head · COCO val2017 | AP: 70.1 -> 70.4 | Improved |
| use ResNet-101 instead of ResNet-50 (ablation) | simple baseline · COCO val2017 | AP: 70.4 -> 71.4 | Improved |
| use ResNet-152 instead of ResNet-50 (ablation) | simple baseline · COCO val2017 | AP: 70.4 -> 72.0 | Improved |
| increase input size from 128x96 to 256x192 (ablation) | simple baseline ResNet-50 · COCO val2017 | AP: 60.6 -> 70.4 | Improved |
| increase input size from 256x192 to 384x288 (ablation) | simple baseline ResNet-50 · COCO val2017 | AP: 70.4 -> 72.2 | Improved |
| deconvolution head vs CPN with OHKM at same backbone and size | CPN ResNet-50 · COCO val2017 | AP: 69.4 -> 70.4 | Improved |
| add joint propagation with optical flow (R-FCN detector) (ablation) | tracking baseline · PoseTrack val | MOTA: 57.6 -> 61.4 | Improved |
| add joint propagation with optical flow (FPN-DCN detector) (ablation) | tracking baseline · PoseTrack val | MOTA: 59.8 -> 62.1 | Improved |
| add joint propagation with optical flow (ResNet-152) (ablation) | tracking baseline · PoseTrack val | MOTA: 62.0 -> 64.8 | Improved |
| replace bbox similarity with flow-based pose similarity (single frame) (ablation) | tracking with joint propagation · PoseTrack val (FPN-DCN) | MOTA: 62.1 -> 62.4 | Improved |
| use multi-frame flow similarity instead of single-frame (ablation) | tracking with joint propagation · PoseTrack val (FPN-DCN) | MOTA: 62.4 -> 62.9 | Improved |
Closest papers by meaning
- PoseTrack: A Benchmark for Human Pose Estimation and Tracking (CVPR 2018 · similarity 0.84)
- SIMPLE: SIngle-network with Mimicking and Point Learning for Bottom-up Human Pose Estimation (AAAI 2021 · similarity 0.83)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.82)
- Detect-and-Track: Efficient Pose Estimation in Videos (CVPR 2018 · similarity 0.82)
- The Devil Is in the Details: Delving Into Unbiased Data Processing for Human Pose Estimation (CVPR 2020 · similarity 0.82)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.81)
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.
