Disease detection on the plant leaves by deep learning

Abstract

Plant disease detection by using different machine learning techniques is very popular field of study. Many promising results were already obtained but it is still only few real life applications that can make farmer’s life easier. The aim of our research is solving the problem of detection and preventing diseases of agricultural crops. We considered several models to identify the most appropriate deep learning architecture. As a source of the training data, we use the PlantVillage open database. During research, the problem with PlantVillage images collection was detected. The synthetic nature of the collection can seriously affect the accuracy of the neural model while processing real-life images. We collected a special database of the grape leaves consisting of four set of images. Deep siamese convolutional network was developed to solve the problem of the small image databases. Accuracy over 90% was reached in the detection of the Esca, Black rot and Chlorosis diseases on the grape leaves. Comparative results of various models and plants using are presented.