Real-Time Feedback and Benchmark Dataset for Isometric Pose Evaluation

2025-06-13

AI Summary

In home-based isometric training, the absence of professional supervision frequently leads to incorrect postures, exercise-related injuries, and low user engagement. Method: We propose the first end-to-end real-time isometric posture assessment system. Our approach comprises (i) constructing the largest multi-class isometric exercise video dataset to date (3,600+ clips), (ii) designing a three-dimensional evaluation metric integrating classification accuracy, erroneous joint localization, and model confidence, and (iii) deploying a graph neural network–based pose recognition model coupled with real-time video stream analysis and fine-grained motion discrimination. Results: Experiments demonstrate substantial improvements in diagnostic accuracy and robustness for home-based posture assessment. The system enables personalized, real-time feedback for rehabilitation and physical therapy applications, and—critically—provides the first empirical validation of the feasibility and practical utility of end-to-end isometric posture assessment.

Technology Category

• Computer Vision: Video Understanding & Activity Analysis
• Intelligent Robots: State Estimation
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Isometric exercises appeal to individuals seeking convenience, privacy, and minimal dependence on equipments. However, such fitness training is often overdependent on unreliable digital media content instead of expert supervision, introducing serious risks, including incorrect posture, injury, and disengagement due to lack of corrective feedback. To address these challenges, we present a real-time feedback system for assessing isometric poses. Our contributions include the release of the largest multiclass isometric exercise video dataset to date, comprising over 3,600 clips across six poses with correct and incorrect variations. To support robust evaluation, we benchmark state-of-the-art models-including graph-based networks-on this dataset and introduce a novel three-part metric that captures classification accuracy, mistake localization, and model confidence. Our results enhance the feasibility of intelligent and personalized exercise training systems for home workouts. This expert-level diagnosis, delivered directly to the users, also expands the potential applications of these systems to rehabilitation, physiotherapy, and various other fitness disciplines that involve physical motion.

Problem

• Lack of expert supervision in isometric exercise training
• Risks from incorrect posture and lack of feedback
• Need for real-time pose evaluation and correction

Innovation

• Real-time feedback system for isometric poses
• Largest multiclass isometric exercise video dataset
• Novel three-part metric for robust evaluation