Disease Identification using Deep Learning in Agriculture: A Case Study of Cotton Plant

Abstract

Among all the agrician products, cotton is known as “Ready Cash Crop” and it plays the significant role in the stability of the economy of a country. Therefore, it is extremely important to monitor the cotton crop from the numerous diseases. Unfortunately, sometimes human eyes not be able to analyze these diseases at earlier stage and that will affect not only the quality and also the quantity of the cotton crops. To address this early monitoring issue we proposed an interactive framework based on target feature extraction and deep learning model for cotton leaf screening to deal with these well-known dangerous diseases; Grey Mildew, Cercospora, Bacterial Blight and Alternaria. In this study we chosen our own collected dataset that contains 522 images of cotton leaves that were collected from the field (Cotton agricultural areas near the Multan city). The performance evaluation matric indicates the algorithm secure; 85.42% overall accuracy, 0.8542 precision, 0.8542 recall, 0.854 F1 score and 0.817 kappa coefficient indicates the generalization and acceptability of the model. The proposed framework not only assists the agronomist but also the farmer because of early identification of diseases from cotton crop and to avoid from the massive loss. It make better decisions for cotton crop management and contributes in the sustainability of the economy.