Deep Learning and YOLOv10 architecture to detect Cercospora nicotianae and Alternaria alternata diseases in tobacco crops

Abstract

In this research, a mobile application based on machine learning was developed to detect and control diseases caused by Cercospora nicotianae and Alternaria alternata on the leaves of the tobacco plant at the Tabacalfa tobacco company in the Ventanas canton. The application used inductive and deductive methods to gather information from various sources and analyze tobacco leaf image data to train with the Roboflow application and 253 images. The results showed that the model's performance improved with training iterations, with metrics such as recall, mAP50, and precision indicating the model's high ability to accurately identify and classify the types of diseases Alternaria alternata and Cercospora nicotianae in tobacco leaves by images. The model with a higher confidence threshold is considered to have better overall performance, with a higher precision-recall curve. In addition to the application of a satisfaction survey to 70 individuals, which revealed a positive user interface, clear navigation, and optimal functionality for detecting diseases in tobacco plants. In addition, an application of technologies such as deep learning and YOLOv10 is praised for a high satisfaction level of 97.14% among the respondents. In conclusion, it is evident that the mobile application has significant economic benefits, as it can reduce crop losses and improve productivity, working conditions, job satisfaction, and reduce the use of pesticides.