Klasifikasi Citra Gerakan Olahraga Dalam Gym Menggunakan Graph Convolutional Network

Abstract

The participation rate in sports activities among Indonesians remains low, with the Sport Development Index (SDI) in 2022 recording only 30.93%, a decline from 32.80% in the previous year. Simple kind of sport can be followed is gym. This study aims to introduce and promote basic gym movements such as bench press, squat, and deadlift to encourage greater engagement in sports activities. This research utilizes Deep Learning technology based on Graph Convolutional Network (GCN) to classify gym movement images into three classes: benchpress, squat, and deadlift. The study focuses on comparing various hyperparameters, including model type, batch size, and dropout, to determine the optimal configuration with the best performance.The results indicate that the GCN model achieved an F1 Score of 0.8667, demonstrating strong performance in classifying gym movement images. A simple web-based application was developed as an implementation to facilitate automatic gym movement classification.