Tomato Leaf Disease Detection using Convolution Neural Network (CNN)

Abstract

Leaf diseases are the major problem in the agricultural sector, which affects crop production as well as economic profit. In certain cases, these diseases may not destroy the plant, but they do reduce yield and quality significantly.  Therefore, the main purpose of this research is to identify the tomato leaf disease using image processing technique, to develop an automatic disease detection system for tomato leaf and to classify tomato leaf disease using Convolution Neural Network (CNN).  The method to be used in this research consist of three-part which are pre-processing, segmentation, feature extraction and classification using Convolutional Neural Network (CNN).  The experimental results have shown that the proposed method is able to successfully detect the leaf disease of affected tomato leaf. The average accuracy for Healthy Leaf is 93.6769 %, sensitivity is 93.3591 % and Fl-Score is 93.0463%. While, for Late Blight Leaf the accuracy is 92.0857%, sensitivity is 92.4698% and Fl-Score is 92.1244%. Meantime, for Septoria Spot Leaf the accuracy 91.3381%, sensitivity is 91.1303% and Fl-Score is 91.0449%. As the result show the healthy leaf has high accuracy, sensitivity and f1-score because healthy leaf has least affected area among three datasets. In future, it is fair to propose expanding the present model to identify other diseases in different plants. It may be possible to investigate the severity of the disease's spread using an automated approach, which could provide additional assistance to persons dealing with the diseases.