Evaluation of Convolutional Neural Networks for Coffee Leaf Rust Classification

Coffee is a beverage present in the lives of many people worldwide and is of great importance to the economies of various countries. Coffee leaf rust is a serious disease that affects crops worldwide, and identifying it quickly and accurately helps in its control. This paper uses a publicly available image dataset to evaluate the performance of Convolutional Neural Networks (CNNs) in the context of the automatic classification of rust on coffee leaves considering binary and multi-class classification. Among the evaluated networks, ResNet achieved the best results, with an accuracy of 95.19% for binary classification and 78.03% for multi-class classification. This study contributes to the application of deep learning as a tool for farmers, enabling the early detection of rust on coffee leaves and aiding in decision-making related to crop management.