Image classification of beef and pork using You Only Look Once (YOLO) method

Abstract

The fraudulent act of mixing pork with beef often causes concern for consumers in Indonesia about the authenticity of the meat products purchased. This study used the YOLO algorithm to classify pork and beef images to address the problem. The method used in this study was first a dataset of pork and beef images collected and labeled with information about the type of meat. After passing through pre-processing, the YOLOv5 algorithm is trained and validated using the dataset. The evaluation results showed excellent performance with an average precision level of 97.9%, 100% recall, and 97.2% mAP50-95 on validation data. When tested, the algorithm was able to recognize and distinguish meat types with varying confidence levels, although some predictions had low confidence levels that required improved performance. The YOLO algorithm succeeded in obtaining an effective model in detecting and classifying objects on meat images, especially in the identification of beef and pork.