This paper presents a comprehensive investigation into the mathematical modeling and statistical analysis of classification of mango leaf diseases employing state-of-the-art Deep Learning models. The literature review underscores the significance of automated disease detection in agriculture, with Convolutional Neural Networks (CNNs) emerging as pivotal tools for image-based tasks. The proposed methodology encompasses meticulous preprocessing, optimal hyperparameter tuning, and fine-tuning of pretrained models (Inception V3, MobileNet V3 Small, MobileNet V3 Large, and ResNet50). Results indicate rapid convergence and outstanding accuracy during initial training, with all models achieving 100% accuracy on both validation and test datasets. K-Fold cross-validation affirms the models’ consistency, with Inception V3 demonstrating leading performance. Detailed analyses, including training and loss graphs and confusion matrices, offer nuanced insights and highlight areas for refinement, particularly in distinguishing Healthy leaf, Gall Midge, and Anthracnose. This research positions the proposed methodology as a promising approach with potential applications in real-world agricultural scenarios, where precise disease detection is critical for effective crop management and optimal yields.