Coffee Bean Classification Using Deep Learning

Abstract—Coffee bean classification, an integral process in the coffee industry, has traditionally relied on manual methods. This study uses deep learning and computer vision techniques to classify coffee beans in a novel way. The model makes use of transfer learning and a ResNet50-based architecture that has been adjusted for the differentiation of coffee beans according to their visual characteristics. Approach is equipped with data augmentation and re-scaling, contributing to improved generalization and model robustness.

The study includes the creation and assessment of a strong classification model capable of identifying the different characteristics of post-roasted coffee beans, attaining a remarkable accuracy rate of 97% . Extensive testing shows how well the model recognizes the visual properties of coffee beans.