KLASIFIKASI AKTIVITAS OLAHRAGA BERDASARKAN CITRA FOTO DENGAN MENGGUNAKAN METODE CONVOLUTIONAL NEURAL NETWORK

Abstract

In an era of advancing technology and information, sports are also receiving increasing attention from various sectors, including enthusiasts and participants in the sports industry. However, to better understand and manage the sports world, a thorough analysis and understanding of various aspects of sports are necessary, including classification and recognition of different types of sports. One potent and effective approach to image pattern recognition is the Convolutional Neural Network (CNN). CNN is a classification method particularly suitable for classifying digital images. The architecture of CNN is designed effectively to recognize objects within images. The dataset employed comprises 2348 samples for training, 294 samples for testing, and 294 samples for validation. The training process of the CNN model using DenseNet121 architecture yields an accuracy rate of 99%, with a validation accuracy rate of 88.78%. Through this research, it is expected that the application of CNN will create a system capable of automatically and accurately identifying the types of sports being performed by individuals or groups based on images or captured visuals of sporting activities.