Classification of Archery Poses Using YOLO And LSTM Methods

Abstract

The classification of archery movements presents a significant challenge in the development of technology to support athlete training. This study aims to develop an archery movement classification system using a combination of YOLO for pose detection and Long Short-Term Memory (LSTM) for temporal classification. The system processes archery training videos into a sequence of images. The aim of this research is to train LSTM model to recognize patterns from four predefined archery movement classes: stand, extend, hold, and release. Evaluation results show that the system achieved an accuracy of 64.47%. Furthermore, analysis using precision, recall, and F1-score metrics indicates varying performance across the movement classes, with the highest F1-score of 83.75% achieved in the release class. This study contributes to the development of machine learning-based technology to support sports training, particularly in archery, by offering a data-driven approach capable of automatically recognizing and evaluating athletes' movements.