Detection of Freestyle Swimming Technique Mistakes Using Computer Vision Technology: A Case Study of Students Faculty Sports Science

Abstract

This study aims to analysed front crawl stroke technical mistakes utilizing computer vision technology. Conventional evaluation of swimming technique still faces various limitations, such as the subjectivity of assessment and the difficulty of observing movements in detail. Therefore, the application of computer vision technology is expected to provide a more objective and accurate analysis of movement. This study employs a quantitative approach using a descriptive-evaluative method and a cross-sectional study design. The research sample consisted of 300 students from students at the Faculty of Sports Science, Medan State University, comprising 246 male students and 54 female students. The dataset consists of video recordings of students swimming the freestyle stroke. Afterwards, the videos were analyzed using a computer vision system based on pose estimation. Data analysis was conducted using descriptive and inferential statistics to identify types of technical mistakes and measure the system's accuracy level. The results of the study show that the most common technical error was unsynchronized arm and leg coordination at 32.3%, followed by lifting the head too high while breathing at 30.7%, irregular leg kicks at 28.3%, a hip position that is too low at 27.0%, excessive head rotation at 26.0%, insufficient arm pull at 24.7%, and an imperfect arm push at 23.0%. The system testing results showed an accuracy of 90.8%, precision of 89.5%, recall of 90.1%, and an F1-score of 89.8%. In addition, the Intraclass Correlation Coefficient (ICC) was 0.87. This indicates a high level of agreement between the system's detection results and the experts' assessments. Therefore, computer vision technology serves as an effective alternative tool for evaluating freestyle swimming learning in higher education institutions.