Deep Learning Models for Cotton Leaf Disease Detection with VGG-16

Naeem A.B.; Senapati B.; Chauhan A.S.; Kumar S.; Gavilan J.C.O.; Abdel-Rehim W.M.F. · International Journal of Intelligent Systems and Applications in Engineering (2023)

Resumen

Cotton, potatoes, tomatoes, and chilies are the principal crops exported from Pakistan. Cotton is produced in greater quantities in India, China, and the US than in Pakistan. Eighty percent of all public oil is made up of cotton. Like other crops, it is being attacked by viruses, which lowers production and reduces economic revenue. There is misinformation on how illness diagnosis and care affect a region's ability to produce more. In this research, deep convolutional neural networks (CNN) were trained to recognize three forms of cotton leaf disease using a transform learning approach (Cotton leaf curl virus, fusarium wilt, bacterial blight). This study's main goal was to develop a single framework that could handle the challenging process of finding, identifying, and diagnosing cotton leaf disease. Additionally, to improve their performance on datasets of healthy and allergic cotton leaves, bigger weight parameter optimization using the Adam and RMSProp optimizers was explored. When compared to the other DL meta-architectures, Inception-VGG-16 trained with the feature extractor showed the greatest mean average accuracy. The recommended method was found to be new since it distinguished between leaf types that were healthy and those that were unhealthy. Using the DL approach to accurately identify cotton leaf disease would help to avoid the adverse effects of dietary management issues. On photos of cotton leaves, the trained model recognizes and labels the four classes (Cotton leaf curl virus, bacterial blight, fusarium wilt, healthy). CNN was 98% accurate overall.

Instituciones: Universidad Privada del Norte