Implementation of Green Coffee Bean Quality Classification Using Slim-CNN in Edge Computing

Abstract

As one of the most important economic industries, how to improve the quality and output of the coffee industry is important. Defective coffee beans affect the flavor of coffee after roasting and grinding. In order to reduce the cost of labor and time, it is effective to use a convolutional neural network (CNN) model to identify defective green coffee beans. However, the complexity and huge parameters of the CNN model make the edge computing devices spend too much time on identification. Therefore, we introduced a lightweight deep learning network Slim-CNN to classify green coffee beans. Experiment results show that Slim-CNN achieves 92% accuracy with 6 times fewer parameters than MobileNet and 270 times fewer parameters than VGG16. The Slim-CNN model can be used on different edge computing devices to reduce labor costs in the coffee industry and improve the quality of coffee.