Research map / 3D vision / Human pose estimation
Learnable Triangulation of Human Pose
Karim Iskakov, Egor Burkov, Victor Lempitsky, Yury Malkov
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 learned camera-joint confidences to algebraic triangulation (ablation) | algebraic triangulation · Human3.6M (Protocol 1, relative to pelvis, average) | MPJPE (mm): 26.9 -> 22.6 | Improved |
| add learned confidences to algebraic triangulation (ablation) | algebraic triangulation · CMU Panoptic validation (4 cameras) | MPJPE (mm): 33.4 -> 21.3 | Improved |
| add learned confidences to algebraic triangulation (ablation) | algebraic triangulation · Human3.6M (absolute positions, filtered validation, average) | MPJPE (mm): 24.5 -> 19.2 | Improved |
| use volumetric triangulation instead of algebraic | triangulation · Human3.6M (Protocol 1, relative to pelvis, average) | MPJPE (mm): 22.6 -> 20.8 | Improved |
| use volumetric triangulation instead of algebraic | triangulation · CMU Panoptic validation (4 cameras) | MPJPE (mm): 21.3 -> 13.7 | Improved |
| use volumetric triangulation instead of algebraic | triangulation · Human3.6M (absolute positions, filtered validation, average) | MPJPE (mm): 19.2 -> 17.7 | Improved |
| use algebraic triangulation with confidences instead of RANSAC | triangulation · Human3.6M (Protocol 1, relative to pelvis, average) | MPJPE (mm): 27.4 -> 22.6 | Improved |
| use softmax aggregation instead of sum aggregation of view volumes (ablation) | volumetric triangulation · Human3.6M (Protocol 1, relative to pelvis, average) | MPJPE (mm): 21.3 -> 20.8 | Improved |
| use softmax aggregation instead of confidence aggregation (ablation) | volumetric triangulation · CMU Panoptic validation (4 cameras) | MPJPE (mm): 14.0 -> 13.7 | No clear effect |
| add heatmap regularizer (beta 0.01) to volumetric L1 loss (ablation) | volumetric triangulation · Human3.6M (Protocol 1, relative to pelvis, average) | MPJPE (mm) | No clear effect |
| use algebraic soft-MSE loss instead of L1 in volumetric triangulation (ablation) | volumetric triangulation · Human3.6M (Protocol 1, relative to pelvis, average) | MPJPE (mm) | Hurt |
| train on CMU and test on Human3.6M instead of training on Human3.6M | volumetric triangulation · CMU-trained model on Human3.6M (elbows, wrists, knees) | MPJPE (mm): 16-18 -> 34 | Hurt |
Closest papers by meaning
- Generalizable Human Pose Triangulation (CVPR 2022 · similarity 0.91)
- Probabilistic Triangulation for Uncalibrated Multi-View 3D Human Pose Estimation (ICCV 2023 · similarity 0.90)
- MetaPose: Fast 3D Pose From Multiple Views Without 3D Supervision (CVPR 2022 · similarity 0.90)
- A Simple yet Effective Baseline for 3D Human Pose Estimation (ICCV 2017 · similarity 0.90)
- MV-SSM: Multi-View State Space Modeling for 3D Human Pose Estimation (CVPR 2025 · similarity 0.89)
- SelfPose3d: Self-Supervised Multi-Person Multi-View 3d Pose Estimation (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.
