<![CDATA[Coffee bean quality is a crucial factor that determines the commercial value and overall quality of coffee products. In practice, coffee bean quality assessment is still largely performed manually through visual inspection, which may lead to subjectivity and inconsistent results. Therefore, this study aims to apply a Convolutional Neural Network (CNN) method to automatically classify coffee bean quality based on visual images. This study employs a quantitative approach using a computational experimental method. The dataset consists of 90 coffee bean images categorized into three quality classes, namely good, medium, and poor, with 30 images in each class. All images undergo preprocessing stages, including image resizing to 150 × 150 pixels, pixel value normalization, and data augmentation. The CNN model is developed using the TensorFlow and Keras frameworks on the Google Colab platform and trained using the Adam optimizer and categorical crossentropy loss function with an optimal number of 20 epochs. The results show that the CNN model is able to learn basic visual patterns of coffee beans progressively, as indicated by an increase in accuracy and a decrease in loss values during the training process. However, evaluation using a confusion matrix reveals that the model still faces difficulties in distinguishing between quality classes with similar visual characteristics, particularly between the good, medium, and poor categories. Nevertheless, this study demonstrates that CNN has strong potential as an initial approach for developing an automated system to support objective and consistent coffee bean quality classification.]]>
