Research map / 3D vision / Human pose estimation
DeciWatch: A Simple Baseline for 10× Efficient 2D and 3D Pose Estimation
Ailing Zeng, Xuan Ju, Lei Yang, Ruiyuan Gao, Xizhou Zhu, Bo Dai, Qiang Xu
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 DenoiseNet (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 79.8 -> 77.2 | Improved |
| use Transformer DenoiseNet instead of TCNs (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 80.5 -> 77.2 | Improved |
| use Transformer DenoiseNet instead of MLPs (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 79.5 -> 77.2 | Improved |
| use Transformer RecoverNet instead of linear interpolation (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 79.8 -> 77.2 | Improved |
| use Transformer RecoverNet instead of TCNs (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 172.3 -> 77.2 | Improved |
| add MLPs at the last layer of TCN recovery (ablation) | TCN RecoverNet · 3DPW (PARE) | MPJPE: 172.3 -> 99.5 | Improved |
| use Transformer RecoverNet instead of MLPs (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 78.0 -> 77.2 | Improved |
| lower sampling ratio from 100% to about 20% (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 78.9 -> 75.0 | Improved |
| lower sampling ratio below 8% (N above 12) (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 75.0 -> 85.3 | Hurt |
| vary input window size 11 to 201 (ablation) | DeciWatch · 3DPW | MPJPE | No clear effect |
| vary loss weight lambda from 1 to 10 (ablation) | DeciWatch · 3DPW (PARE) | MPJPE: 78.0 -> 77.2 | No clear effect |
| use DeciWatch recovery instead of cubic spline at interval 10 | DeciWatch · Human3.6M labeling | MPJPE: 5.36 -> 2.89 | Improved |
Later papers that compared with this method
1 later accepted paper reports DeciWatch as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- MixSynthFormer: A Transformer Encoder-like Structure with Mixed Synthetic Self-attention for Efficient Human Pose Estimation (ICCV 2023 · better on 8 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DynPose: Largely Improving the Efficiency of Human Pose Estimation by a Simple Dynamic Framework (CVPR 2025 · similarity 0.84)
- Robust and Accurate Line- and/or Point-Based Pose Estimation without Manhattan Assumptions (ECCV 2016 · similarity 0.84)
- 3D Human Pose Estimation With Spatio-Temporal Criss-Cross Attention (CVPR 2023 · similarity 0.84)
- Integration of Probabilistic Pose Estimates from Multiple Views (ECCV 2016 · similarity 0.84)
- FreeMan: Towards Benchmarking 3D Human Pose Estimation under Real-World Conditions (CVPR 2024 · similarity 0.82)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.82)
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.
