Mango Leaf Disease Severity Scale Classification Using ResNext50 Model: A Deep Learning-Based Approach

The production of mangoes has a serious obstacle in the form of mango leaf diseases, which have a negative influence not only on yield but also on the health of the trees. The mango leaf sooty mold disease stands out among these illnesses due to the negative effects that it has on photosynthesis as well as the overall vigor of the tree. The purpose of this research study is to investigate the application of the Residual Network with 50 layers and the next-generation architecture (ResNext50) deep learning (DL) model for the categorization of the severity of mango leaf sooty mold disease. The research reveals that the model is capable of performing automated disease severity assessment by using a dataset consisting of 25,000 photos collected from a variety of mango fields. An overall accuracy of 94.61% is achieved by the ResNext50 architecture, with layer-wise parameter analyses, performance metrics, and confusion matrices providing insights into its categorization capabilities. Comparisons made both between models and within models highlight the competitive advantage that the model has within the field. This research not only demonstrates the utility of DL for disease management but also paves the way for further improvements to be made to practical applications in agriculture. It also points to a bright future for the automated evaluation of mango leaf diseases.