Research map / 3D vision / Human pose estimation
Conservative Wasserstein Training for Pose Estimation
Xiaofeng Liu, Yang Zou, Tong Che, Peng Ding, Ping Jia, Jane You, B.V.K. Vijaya Kumar
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 |
|---|---|---|---|
| use squared arc length instead of arc length as ground metric (ablation) | Wasserstein loss, one-hot · CAVIAR head pose | MAAD: 18.8 -> 17.1 | Improved |
| use concave chord-length ground metric instead of arc length (ablation) | Wasserstein loss, one-hot · CAVIAR head pose | MAAD: 18.8 -> 19.1 | Hurt |
| add adaptive ground metric learning (ablation) | Wasserstein arc length · CAVIAR head pose | MAAD: 18.8 -> 17.5 | Improved |
| add adaptive ground metric learning (ablation) | Wasserstein squared arc length · CAVIAR head pose | MAAD: 17.1 -> 17.3 | No clear effect |
| use exact instead of approximate Wasserstein computation (ablation) | Wasserstein squared arc length · CAVIAR head pose | MAAD: 17.8 -> 17.1 | Improved |
| add adaptive ground metric learning (ablation) | Wasserstein arc length · TUD pedestrian (36 poses) | Mean AE: 19.1 -> 20.5 | Hurt |
| add Binomial-uniform conservative target label (ablation) | Wasserstein squared arc length · TUD pedestrian (36 poses) | Mean AE: 18.5 -> 16.3 | Improved |
| use Poisson instead of Binomial conservative label (ablation) | Wasserstein squared arc length · TUD pedestrian (36 poses) | Mean AE: 16.3 -> 17.7 | Hurt |
| add conservative target label (ablation) | Wasserstein arc length · EPFL vehicle | Mean AE: 6.46 -> 6.30 | Improved |
| add conservative target label (ablation) | Wasserstein squared arc length · EPFL vehicle | Mean AE: 6.23 -> 6.04 | Improved |
| use cube instead of square of arc length as ground metric (ablation) | Wasserstein loss, one-hot · EPFL vehicle | Mean AE: 6.23 -> 6.47 | Hurt |
| use Huber ground metric instead of squared arc length (ablation) | Wasserstein loss, one-hot · EPFL vehicle | Mean AE: 6.23 -> 6.20 | No clear effect |
Closest papers by meaning
- Augmented Sliced Wasserstein Distances (ICLR 2022 · similarity 0.87)
- Wasserstein of Wasserstein Loss for Learning Generative Models (ICML 2019 · similarity 0.86)
- Learning Wasserstein Embeddings (ICLR 2018 · similarity 0.86)
- Sliced Wasserstein Estimation with Control Variates (ICLR 2024 · similarity 0.84)
- A New Robust Partial p-Wasserstein-Based Metric for Comparing Distributions (ICML 2024 · similarity 0.84)
- Statistical and Computational Guarantees of Kernel Max-Sliced Wasserstein Distances (ICML 2025 · similarity 0.84)
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.
