Transfer Learning Approach for Classification of Beef Meat Regions with CNN

Abstract

Accurate identification of beef components is crucial for the meat industry, encompassing consumer confidence, food safety, and quality control. This study addresses the challenge by developing a robust model for beef component classification using RGB images obtained from smartphones. A diverse dataset was collected outside a controlled laboratory environment, closely resembling real-world conditions. Three CNN-based models, EfficientNetV2S, ResNet101, and VGG16, were fine-tuned and evaluated on the dataset. The results demonstrated the effectiveness of the models in accurately classifying beef components. EfficientNetV2S achieved the highest performance, with precision, recall, and F1-score values of 0.92 for all classes. This research bridges the gap between non-destructive detection technologies and end users, providing a practical and reliable solution for beef component identification in various applications. © 2023 IEEE.