Shifting the Weight: Applications of AI in Olympic Weightlifting

The role of humans in time-pressured decision-making processes within sports has been critically examined in psychological research. This is particularly relevant in complex movement sports such as Dressage, Gymnastics, and Olympic Weightlifting. Not only are humans susceptible to bias, but they also lack the necessary processing capacity to assess intricate movements in real-time. Although some research has been conducted in this space very few use Computer Vision based approaches. To address this issue, this research proposes a novel Computer Vision solution to automate the judging process in Olympic Weightlifting. The solution incorporates LSTM-based Gesture Recognition and Human Pose Estimation using Mediapipe. The feasibility and effectiveness of the proposed solution are assessed by leveraging a combination of videos from the official Olympics YouTube channel and amateur recorded videos captured from the perspective of the Olympic Weightlifting Centre judge. The findings indicate a high degree of success in achieving the research objective. The solution achieved a validation accuracy of 96% and an average F1 score of 0.9l. These results demonstrate the plausibility and efficacy of the proposed approach in automating the judging process within Olympic Weightlifting. By automating this process, the potential influence of human bias can be mitigated while improving the real-time assessment of complex movements. The implications of these findings extend beyond Olympic Weightlifting and have the potential to enhance judging processes in other complex movement sports as well.