This paper presents a smartphone-based diagnostic system for cassava crop health that uses machine learning to solve the problem of identifying disease in the field from analysis of plant leaf images. It deals with two important items of information: disease incidence and disease severity. This application provides a classification system that can determine the state of disease in a plant. Five classes can be represented – one healthy class and four disease classes. This paper presents ways of extracting different features from leaf images and shows how they can affect the performance of the classifier. The smartphone-based system can provide real-time prediction of the state of health of a farmer’s plot: the farmer is able to upload an image of a plant and obtain a disease score from a remote server.