Implementasi YOLOv5 untuk Deteksi Objek Mesin EDC: Evaluasi dan Analisis

Abstract

The Electronic Data Capture (EDC) machine is essential for facilitating non-cash transactions, yet its efficient detection remains a challenge. This study explores the implementation of the You Only Look Once (YOLOv5) algorithm to enhance EDC machine detection. The objective is to improve accuracy and efficiency in detecting EDC machines in various environments, thereby enhancing transaction security and efficiency. The research methodology involved acquiring a diverse dataset from social media platforms and the internet, comprising 396 images after augmentation. Using Roboflow, the dataset was annotated and divided into training, validation, and testing sets. The YOLOv5 model was trained on Google Colab, achieving a Precision of 97.1%, Recall of 86.4%, and mean Average Precision (mAP50) of 92.0% on the validation set. The results demonstrate that YOLOv5 effectively detects EDC machines with high accuracy across different scenarios, validating its robustness in real-world applications. This research suggests that YOLOv5 can significantly improve transaction security and efficiency in retail and service industries. The implications of this research are substantial for industry stakeholders and decision-makers, offering a reliable solution to enhance transaction security and streamline non-cash payment processes. By integrating YOLOv5, businesses can optimize operational efficiency and customer service, paving the way for broader adoption of advanced computer vision technologies in commercial applications