Mangoes (Mangiferaindica) are vulnerable to bacterial and fungal infections as well as abiotic stresses, necessitating early and precise disease detection. This study evaluates multiple machine learning and deep learning models, integrating advanced optimization techniques to enhance classification performance. A range of AI approaches was employed, including convolutional neural networks (CNNs), fuzzy logic, and swarm intelligence-based optimization. Among the tested models, ResNet50 achieved the highest accuracy (99.7%), surpassing VGG-16 (94%) and EfficientNetV2-B0 (97.13%). The MCNN model, combined with CNN-Fuzzy and SA-GSO optimization, also attained 97.13%, while YOLOv3 demonstrated strong real-time detection capabilities with 83.33%. Additionally, traditional models such as feedforward neural networks (FFNN, 93%), k-nearest neighbors (KNN, 91%), and support vector machines (SVM, 64%) were included for comparison. Hybrid optimization techniques like particle swarm optimization (PSO) and image quality metrics (PSNR, MSE) further improved classification performance, presenting an efficient and comprehensive strategy for mango disease management.