Towards Sustainable Agriculture: Mango Leaf Disease Classification with Deep Learning Models

This work proposes a deep learning-based method using the EfficientNetB5 model for the classification of mango leaf images into eight categories, including six disease classes (Powdery Mildew, Cutting Weevil, Anthracnose, Bacterial Canker, Sooty Mould, and Gall Midge), a healthy class, and Die Back. Training, validation, and test sets were formed from a publicly available collection comprising 4,000 labeled images. Using Adamax optimizer with hyperparameters including a batch size of 32, a learning rate of 0.001, and a dropout rate of 0.2, the model was fine-tuned. Supported by high precision, recall, and F1-scores across all classes, data augmentation techniques and an input image size of (224,224,3) helped the model, to achieve an amazing test accuracy of 97%. Minimal misclassification revealed by the confusion matrix analysis confirmed the model's consistency in disease diagnosis. This work conforms with the objectives of sustainable agriculture and responsible production through exact disease identification, so enhancing food security and supporting economic development in agricultural communities. For more general relevance in agricultural activities, future developments could be real-time deployment using mobile apps.