Mango Diseases Classification Using Deep Learning

This paper presents a mango illness detection model that uses the EfficientNetB0 design, which was chosen because it is both fast and accurate. The structure of the model is made up of layers that can be changed using different hyperparameters. Some of these levels are dropout, batch normalization, and thick. A batch size of 32 was picked to boost computation speed, and the model was trained across 15 epochs. Adamax, an adaptive optimization method, was used with a learning rate of 0.001 to find the best mix between speed and safety. Due to its speed in multiclass classification tasks, categorical cross-entropy was picked as the loss function during training. The model attained an incredible accuracy of 100% across training, validation, and test datasets, exhibiting consistent performance. The confusion matrix indicated flawless categorization across all classes, demonstrating no misclassifications. Loss measurements similarly demonstrated high convergence, with total losses around 0.20 across datasets. These results emphasize the model's reliability and accuracy in real-world mango disease detection tasks, giving a very promising option for farm disease control.