Effective Machine Learning Model for Disease Recognition and Categorization of Mango Leaves

This is the authors' abstract. We don't add key points for this paper.

Maintaining the health and production of mango farms depends on the early detection of illnesses in mango trees. Despite the vital role agriculture plays, it remains an industry receiving limited attention from the machine-learning community. A noteworthy challenge is the lack of standardized and openly accessible agricultural datasets, impeding researchers from fully leveraging the potential of advanced computational prediction tools. The study focuses on employing deep learning and image processing techniques for diagnosing prevalent mango leaf diseases in India. The goal is to enhance both the quality and quantity of agricultural production by implementing effective disease control measures upon identifying plant diseases. The automation of plant disease detection proves particularly beneficial, reducing the monitoring efforts required by large farms. Rapid identification of leaf diseases is crucial, given that leaves serve as a plant's primary source of nutrition. Notably, none of the seven different mango leaf illnesses that have been observed in India have ever been the subject of earlier research. In this study, we provide a lightweight Resnet50 model that can reliably classify healthy mango leaves and the seven discovered illnesses on mango leaves. The performance of our proposed lightweight Resnet-50 model is evaluated against many pre-trained models, such as Exception, Resnet101, and VGG16. The findings reveal that the lightweight Resnet-50 model achieves the highest testing accuracy, reaching an impressive 95%. This underscores the model's effectiveness in automating the identification of mango leaf diseases, offering a promising solution for improving agricultural practices in India.