Detection of the Precision of the Basic Movement of the Line using the Convolutional Neural Network (CNN)

Abstract

Payeti Christian Junior High School in Kambera District, East Sumba Regency, NTT, still makes marching exercises (PBB) an important part of extracurricular activities and character development of students. However, the evaluation of the basic movements of the United Nations is still carried out manually by observing the movements of the students one by one. This method is inefficient and tends to be subjective. Common mistakes found include not standing upright, head down, misaligned legs, and hands that are not straight when in perfect posture. This study aims to develop an automatic detection system using Convolutional Neural Network (CNN) to evaluate the accuracy of UN basic movements through static imagery. This system is expected to provide a faster, objective, and documented evaluation. The implementation results show a very high level of accuracy, with model confidence above 92% to 99.99%. Out of 200 test data, only one was misclassified. The accuracy and loss graphs show the stability of the model without overfitting, with a final validation accuracy of 98.44%