![[2602.13901] RPGD: RANSAC-P3P Gradient Descent for Extrinsic Calibration in 3D Human Pose Estimation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13901] RPGD: RANSAC-P3P Gradient Descent for Extrinsic Calibration in 3D Human Pose Estimation
Summary
The paper presents RPGD, a novel framework for extrinsic calibration in 3D human pose estimation, combining RANSAC-P3P and gradient descent for improved accuracy in aligning motion capture data with RGB cameras.
Why It Matters
Accurate extrinsic calibration is crucial for applications in computer vision and robotics, particularly in human motion analysis. RPGD offers a robust, automated solution that can enhance the reliability of 3D human pose estimation, which is vital for advancements in fields like augmented reality and human-computer interaction.
Key Takeaways
- RPGD combines RANSAC-P3P and gradient descent for effective calibration.
- The framework is tailored for human poses, enhancing robustness.
- Achieves sub-pixel MPJPE reprojection error in challenging conditions.
- Evaluated on large-scale datasets, demonstrating practical applicability.
- Provides a solution for reliable data collection in 3D human pose estimation.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.13901 (cs) [Submitted on 14 Feb 2026] Title:RPGD: RANSAC-P3P Gradient Descent for Extrinsic Calibration in 3D Human Pose Estimation Authors:Zhanyu Tuo View a PDF of the paper titled RPGD: RANSAC-P3P Gradient Descent for Extrinsic Calibration in 3D Human Pose Estimation, by Zhanyu Tuo View PDF HTML (experimental) Abstract:In this paper, we propose RPGD (RANSAC-P3P Gradient Descent), a human-pose-driven extrinsic calibration framework that robustly aligns MoCap-based 3D skeletal data with monocular or multi-view RGB cameras using only natural human motion. RPGD formulates extrinsic calibration as a coarse-to-fine problem tailored to human poses, combining the global robustness of RANSAC-P3P with Gradient-Descent-based refinement. We evaluate RPGD on three large-scale public 3D HPE datasets as well as on a self-collected in-the-wild dataset. Experimental results demonstrate that RPGD consistently recovers extrinsic parameters with accuracy comparable to the provided ground truth, achieving sub-pixel MPJPE reprojection error even in challenging, noisy settings. These results indicate that RPGD provides a practical and automatic solution for reliable extrinsic calibration of large-scale 3D HPE dataset collection. Comments: Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Robotics (cs.RO) Cite as: arXiv:2602.13901 [cs.CV] (or arXiv:2602.139...