Sesame Seed Disease Detection Using Image Classification

Abstract

Plant diseases cause significant losses in production, economy, quality, and quantity of agricultural products. The economy of Sudan is highly dependent on agriculture. The international demand for Sudanese sesame seeds was remarkably stable throughout 2018. However, the agriculture sector’s contribution to GDP (Gross Domestic Product) growth rate has decreased from 4% in 2018 to 1.2% in 2019. One of the main contributors to the decrease in the contribution of the GDP was diseases. The early detection of diseases is vital in agriculture for efficient crop yield. One of the leading technologies used worldwide for disease detection is machine learning and Specifically Convolutional Neural Networks, which classify images of diseased plants/leaves from healthy plants/leaves. This research compares a developed CNN model and five other ready-made models; VGG16, VGG19, Resnet50, Resnet101, and Resnet152. The dataset contains 1,695 images of sesame leaves grouped into three classes, two of which are of diseases currently affecting the Sesame in Sudan, and the third is of the healthy leaves. The leaves are photographed from different fields in Gadarif State. The developed model achieved the best result with a training accuracy of 90.77% and testing accuracy of 88.5%. Future work and possible improvements to this model were also discussed.