Reproduce “OpenCapBench: A Benchmark to Bridge Pose Estimation and Biomechanics” on OpenResearch

About this paper

Yoni Gozlan, Antoine Falisse, Scott Uhlrich, Anthony Gatti, Michael Black, Akshay Chaudhari

Researchers from Stanford University and the Max Planck Institute for Intelligent Systems developed OpenCapBench, a unified benchmark for evaluating human pose estimation models against biomechanically relevant kinematic metrics. They also introduced SynthPose, a method that uses synthetic data to fine-tune 2D pose estimators for predicting dense, anatomically precise keypoints, demonstrating substantial reductions in joint angle errors for biomechanical analysis.

Key insights

• Biomechanical evaluation metrics, specifically joint angle RMSE, provide more nuanced and diagnostic insights into model performance and training data efficacy compared to general computer vision metrics like PCK.
• Directly predicting dense, anatomically meaningful keypoints from images (via SynthPose) is significantly more effective for accurate biomechanical analysis than inferring SMPL parameters and then projecting vertices.
• Strategic use of diverse synthetic datasets, especially those focused on specific exercises like Infinity VisionFit, is crucial for improving model performance on biomechanically relevant kinematic metrics.

Summary by alphaXiv.