Detection and classification of plant diseases in leaves through machine learning

Abstract

Plant diseases cause significant productivity and economic losses, as well as a reduction in agricultural product quality and quantity. One principal impact on low crop yield is sickness due to bacteria, virus and fungus It is possible to avoid it by employing plant disease detection and categorization procedures. We used machine learning to detect and classify diseases in plant leaves because it evaluates data from several perspectives and categorizes it into one of several predefined classifications. In this research we create a model for the classification task which is sequential model. We trained a convolutional neural network (CNN) with help of the plant village dataset, which have 55,000 images divided into 39 completely distinct categories of each healthy and effected leaves. We trained data by using Adam optimization technique because it almost constantly plays quicker and higher global minimal convergence in comparison to the alternative optimization techniques. We achieved a validation accuracy of 98.74% using the architecture of CNN containing optimized parameters. CNNs, as can be observed, have a high-stop overall performance, making them surprisingly suitable for computerized identification of plant illnesses using simple plant leaf images. The experiment effects completed are similar with different current strategies in literature.