<![CDATA[Beef freshness assessment is a critical aspect of the food industry, where conventional methods relying on visual inspection and olfaction are subjective and prone to human error. This study aims to develop an automated beef freshness classification system utilizing deep learning to enhance the accuracy and objectivity of quality assessments. A total of 2.800 beef images were sourced from an open research data repository, divided equally into fresh and spoiled classes. Three Convolutional Neural Networks (CNN) architectures with transfer learning VGG16, ResNet50-V2, and Inception-V3 were evaluated. The models were systematically tested using learning rate tuning at 0.01, 0.001, and 0.0001 to optimize training convergence. Evaluation results showed that VGG16 outperformed other models in classifying beef freshness. VGG16 achieved a peak testing accuracy of 99.46% at a learning rate of 0.001. The main contribution of this study is the systematic evaluation of transfer learning architectures to establish an optimal baseline for beef quality assessment. By deploying the best performing model into a web based application, this approach offers a practical, objective, and accessible alternative to conventional manual inspection, enabling rapid early detection of beef freshness to improve food safety.]]>
Copyrights © 2026
