Plant Disease Detection using Artificial Intelligence techniques for Agricultural productivity enhancement in Egypt

In the past few years, the use of artificial intelligence and especially Convolution Neural networks (CNN) for different computer vision tasks has been taking over. One of these tasks is detecting diseases in plants since plant diseases is now a topic of interest as it highly affects the crop yield production. Egypt in fact is one of the main countries that should consider plant disease detection for a productive crop yield since Egypt heavily relies on agriculture as it’s income. Moreover, other countries also rely on the crops exported from Egypt such as oranges, as Egypt is considered one of the largest orange exporters in the world. Therefore, deep learning was stated as the current effective approach in smart agriculture to help tackle the crop yield reduction crisis that are caused due to the neglection of plant diseases and the standard visual inspecting techniques used for detecting plant diseases. In this paper different deep learning Convolution Neural Network approaches are proposed using the new plant diseases dataset which consists of 38 classes of 14 types of healthy and diseased crop leaf images to help detect diseases in plants effectively and efficiently. The first proposed approach is a Convolution Neural Network model consisting of four convolution and pooling layers in addition to two fully connected layers and an output layer with 38 neurons for each of the 38 classes which resulted in an accuracy of 98.36%, loss of 0.0972, 98.35% F1 score, 98.36% recall, 98.39% precision, model size of 12.25mb, training time of 22.82 seconds, and total parameters of 668, 134. While the second approach is a Hybrid ResNet50-InceptionV3 model with an accuracy of 96.20%, a precision of 96.16%, and F1-score of 96.45%.