Evaluating Squat Technique in Pound Fitness through Deep Learning and Human Pose Estimations

AI-supported sports applications, such as fitness trackers, virtual sports, and AI coaches, have become increasingly popular during and after the pandemic. AI coaches can provide customized workout programs tailored to individual needs, while virtual sports applications enable people to exercise at home. Pound Fitness is a sport that combines dance movements with bodyweight exercises using drumsticks, which has gained popularity among the community. The aim of this research is to demonstrate that human pose estimation technology can be used to accurately measure the performance of athletes in Pound Fitness movements. For this study, we collected data through front-facing video recordings of Pound Fitness sessions, involving two groups, each with five individuals over four weeks, to evaluate the progression in their movement accuracy and performance. Group A consisted of 5 individuals who participated in a Pound Fitness exercise once a week, and Group B consisted of 5 different individuals who participated in a Pound Fitness exercise twice a week. Our findings reveal tangible improvements in both groups, indicating a positive correlation between regular training and enhanced performance. Notably, more frequent training yielded consistent progress, whereas less frequent, high-intensity sessions resulted in significant performance leaps for certain individuals. These outcomes suggest a potential advantage in diversifying training frequency and intensity to optimize fitness results