Fake Beef Detection with Machine Learning Technique

Abstract

Increased demand for meat causes diverse concerns, including misuse of sales by offering fake meats. Hence, consumers must be shielded from these cheated sellers. Yet, distinguishing faked meat and quality meat is not easy for regular consumers as, at present, meat identification is done manually using visual identification of human vision. Therefore, in this study, we proposed a concept to minimize the above issue by developing a virtual expert to assist in meat inspection. After extracting and pre-processing the relevant images, the model training was accomplished with the SVM, and CNN approaches. The determination of subjection of this classification process is evaluated using F1-Score and precision. Our model evaluation for pork and beef classification utilizing 20% test data against the five classification models showed that the VGG16 produced the highest accuracy rate of 95.20% with 1200 images. Besides, the best accuracy result demonstrated as (Class, F1-Score, Precision) of (Pork, 98.00%, 98.00%) and (Beef, 98.00%, 98.00%).