Deep Learning Convolution Neural Network to Detect and Classify Tomato Plant Leaf Diseases
Thair A. Salih, Ahmed J. Ali, Mohammed Nabeel Ahmed
Abstract
The tomato crop is an important staple in the market and it is one of the most common crops daily consumed. Plant or crop diseases cause reduction of quality and quantity of the production; therefore detection and classification of these diseases are very necessary. There are many types of diseases that infect tomato plant like (bacterial spot, late blight, sartorial leaf spot, tomato mosaic and yellow curved). Early detection of plant diseases increases production and improves its quality. Currently, intelligent approaches have been widely used to detect and classify these diseases. This approach helps the farmers to identify the types of diseases that infect crop. The main object of the current work is to apply a modern technique to identify and classify the disease. Intelligent technique is based on using convolution neural network (CNN) which is a part of machine learning to obtain an early detection about the situation of plants. CNN method depends on feature extraction (such as color, leaves edge, etc.) from input image and on this basis the decision of classification is done. A Matlab m-file has been used to build the CNN structure. A dataset obtained from plant village has been used for training the network (CNN). The suggested neural network has been applied to classify six types of tomato leaves situation (one healthy and five types of leave plant diseases). The results show that the convolution neural network (CNN) has achieved a classification accuracy of 96.43%. Real images are used to validate the ability of suggested CNN technique for detection and classification, and obtained using a 5-megapixel camera from a real farm because most common diseases which infect the planet are similar.