A Coffee Leaf Disease Detection via Deep Learning Algorithms

Among the many beverages served globally, coffee is certainly one of the most popular and it largely depends on the health status of the coffee plants on the quality of coffee beans. Coffee plants can be affected by different diseases and pests, one of which is coffee leaf disease which constitutes a serious threat to global coffee production. The appropriate and early diagnosis of these diseases is important in a bid to prevent their spread as well as maintain the wellbeing of the coffee leaf. Given that the coffee is an integral component of the global economic structure and diseases especially impact its productivity and quality, therefore, it is important to have an automatic mechanism for detecting coffee plant diseases. This research is focusing on the new deep learning technologies which are developed in the recent past experiences for the identification of coffee leaf diseases. The current trends, and the limiting factors hindering coffee leaf disease identification using deep learning and advanced imaging methods are reviewed. In addition, we applied various deep learning frameworks and validated the results obtained from every architecture. The result of this comparison shows a neural network representing the combination is more effective in diagnosing diseases of coffee leaves. MobileNent, ResNet50, and hybrid models display a detection accuracy in the range of 82% to 99.3%. Based on the analysis, the success rate of the MobileNet model reached 99.93% of the success rate of other deep learning frameworks for coffee leaf disease classification. This shows the power of deep learning algorithms to decide coffee leaf diseases on their early stage.