<![CDATA[Coffee is one of the most widely consumed beverage ingredients in Indonesia and has high economic value to improve the community's economy and as a source of foreign exchange. The roasting process is an important stage in coffee processing because it affects the aroma and flavor of coffee. What is often encountered is that visually determining the level of coffee roasting is often inaccurate and prone to human error. To overcome this problem, this study uses a deep learning approach with a transfer learning method based on MobileNet architecture to classify the level of coffee roasting maturity based on digital images. MobileNet was chosen due to its lightweight and fast architecture, suitable for implementation on mobile devices. This research aims to compare the performance of the model in detecting coffee roasting level automatically, efficiently, and objectively. With this approach, it is expected that coffee enthusiasts and producers can easily recognize the type of coffee roasting, support product quality consistency, and reduce dependence on experts in the roasting process. This study analyzed the performance of the classification model with the results showing excellent performance. The model achieved a total accuracy of 99.50%, with consistently high precision, recall, and f1-score values across all classes, including several classes with perfect scores (1,000). Evaluation using ROC curves and AUC also demonstrated the model's ability to distinguish between the two classes.]]>
Copyrights © 2025
