Research map / 3D vision / Human pose estimation
Probabilistic Monocular 3D Human Pose Estimation With Normalizing Flows
Tom Wehrbein, Marco Rudolph, Bodo Rosenhahn, Bastian Wandt
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 heatmap condition (ablation) | normalizing flow pose model · H36MA | CPS: 137.6 -> 171.0 | Improved |
| add L_HM (ablation) | normalizing flow pose model · H36MA | CPS: 157.3 -> 171.0 | Improved |
| add discriminator loss L_gen (ablation) | normalizing flow pose model · H36MA | CPS: 165.5 -> 171.0 | Improved |
| add discriminator loss L_gen (ablation) | normalizing flow pose model · Human3.6M Protocol 2 worst hypothesis | PMPJPE (mm): 284.1 -> 86.8 | Improved |
| add L_MB multi-best loss (ablation) | normalizing flow pose model · H36MA | MPJPE: 76.0 -> 71.0 | Improved |
| use k=5 in L_MB instead of k=1 (best-of-M) (ablation) | normalizing flow pose model · H36MA | CPS: 167.4 -> 171.0 | Improved |
| use k=5 in L_MB instead of k=10 (ablation) | normalizing flow pose model · H36MA | CPS: 168.3 -> 171.0 | Improved |
| use k=5 in L_MB instead of k=50 (ablation) | normalizing flow pose model · H36MA | CPS: 168.0 -> 171.0 | Improved |
| increase hypotheses from M=1 (z0) to M=200 | normalizing flow pose model · Human3.6M Protocol 1 | MPJPE (mm): 61.8 -> 44.3 | Improved |
| sample 200 hypotheses instead of 5 | Li et al. baseline sampled · Human3.6M Protocol 1 | MPJPE (mm): 74.9 -> 59.6 | Improved |
| use flow posterior sampling instead of noise-on-Gaussian baseline | normalizing flow pose model · H36MA increasing samples | MPJPE | Improved |
| use K-Means over 200 hypotheses with 5 clusters | normalizing flow pose model · Human3.6M Protocol 2 | PMPJPE (mm): 42.6 -> 38.4 | Improved |
Closest papers by meaning
- FMPose3D: monocular 3D pose estimation via flow matching (CVPR 2026 · similarity 0.93)
- Weakly Supervised 3D Human Pose and Shape Reconstruction with Normalizing Flows (ECCV 2020 · similarity 0.90)
- Sparseness Meets Deepness: 3D Human Pose Estimation From Monocular Video (CVPR 2016 · similarity 0.90)
- HuManiFlow: Ancestor-Conditioned Normalising Flows on SO(3) Manifolds for Human Pose and Shape Distribution Estimation (CVPR 2023 · similarity 0.90)
- ProPose: Probabilistic 3D Human Pose Estimation with Instance-Level Distribution and Normalizing Flow (AAAI 2025 · similarity 0.89)
- Normalizing Flows on the Product Space of SO(3) Manifolds for Probabilistic Human Pose Modeling (CVPR 2024 · similarity 0.88)
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.
