Implementasi Analisis Kualitas Pengeringan Buah Kopi Berdasarkan Warna Dan Tekstur Menggunakan Convolutional Neural Network (CNN) Pada Solar Dryer Portable

Coffee bean drying is a critical post-harvest stage that determines the quality of the final product. This study implements an analysis of coffee fruit drying quality based on color and texture using a Convolutional Neural Network (CNN) in a portable solar dryer system. The aim of the research is to develop an automated classification model to determine the dryness level of coffee beans by leveraging deep learning technology. The methodology involves collecting coffee bean images at various dryness levels, classified into three main categories: wet (0), medium (1), and dry (2). The image data is then processed through preprocessing, data augmentation, and feature extraction using the CNN architecture. The model was trained and validated to produce accurate predictions of coffee bean dryness levels. The results demonstrate excellent model performance with an overall accuracy of 96%. Specifically, the wet class achieved perfect performance with precision, recall, and F1-score all at 1.00. The dry class achieved a precision of 1.00, recall of 0.87, and F1-score of 0.93, while the medium class achieved a precision of 0.90, recall of 1.00, and F1-score of 0.94. Out of 170 test samples, the model correctly classified the majority of data, with some misclassifications occurring in the dry category, which was predicted as medium. The conclusion indicates that implementing CNN for classifying coffee bean drying quality based on color and texture analysis can yield satisfactory results. However, to further improve the model’s performance, enhancements in feature extraction methods and more suitable data augmentation techniques are required. This research makes an important contribution to the development of automation technology for the coffee processing industry, particularly in the application of portable solar dryers, which can assist coffee farmers in monitoring drying quality in real-time.