Technology Category
Computer Vision: Biometrics, Face, Gesture & PoseIntelligent Robots: State EstimationMachine Learning: Learning with Manifolds
Polar Coordinate-Based 2D Pose Prior with Neural Distance Field
🤖 AI Summary
To address motion blur, severe occlusion, and poor cross-pose domain generalization in sports videos, this paper proposes a polar-coordinate-driven 2D pose prior-guided framework. We innovatively model joint-to-joint relationships via polar coordinates—separately encoding angular and radial components—and design a non-geodesic distance metric decoupled along these dimensions. A neural distance field (NDF) is introduced to enforce geometric consistency during pose refinement. Additionally, we propose a gradient-driven batch projection data augmentation strategy to mitigate the small-sample bottleneck. Evaluated on a real-world long-jump dataset, our method achieves significant improvements in estimation plausibility, robustness, and physical interpretability across diverse poses—even with minimal annotations—while demonstrating exceptional cross-domain generalization capability.
Technology Category
Application Category
📝 Abstract
Human pose capture is essential for sports analysis, enabling precise evaluation of athletes' movements. While deep learning-based human pose estimation (HPE) models from RGB videos have achieved impressive performance on public datasets, their effectiveness in real-world sports scenarios is often hindered by motion blur, occlusions, and domain shifts across different pose representations. Fine-tuning these models can partially alleviate such challenges but typically requires large-scale annotated data and still struggles to generalize across diverse sports environments. To address these limitations, we propose a 2D pose prior-guided refinement approach based on Neural Distance Fields (NDF). Unlike existing approaches that rely solely on angular representations of human poses, we introduce a polar coordinate-based representation that explicitly incorporates joint connection lengths, enabling a more accurate correction of erroneous pose estimations. Additionally, we define a novel non-geodesic distance metric that separates angular and radial discrepancies, which we demonstrate is better suited for polar representations than traditional geodesic distances. To mitigate data scarcity, we develop a gradient-based batch-projection augmentation strategy, which synthesizes realistic pose samples through iterative refinement. Our method is evaluated on a long jump dataset, demonstrating its ability to improve 2D pose estimation across multiple pose representations, making it robust across different domains. Experimental results show that our approach enhances pose plausibility while requiring only limited training data. Code is available at: https://github.com/QGAN2019/polar-NDF.
Problem
Research questions and friction points this paper is trying to address.
Addressing motion blur and occlusions in sports pose estimation
Improving pose accuracy with polar coordinate-based representation
Reducing data needs via gradient-based pose augmentation
Innovation
Methods, ideas, or system contributions that make the work stand out.
Polar coordinate-based 2D pose representation
Non-geodesic distance metric for discrepancies
Gradient-based batch-projection augmentation strategy
🔎 Similar Papers
No similar papers found.
