Artificial intelligence-based mango leaf disease detection: current progress, emerging technologies, challenges, and future perspectives

  Full text PDF

Mango (Mangifera indica L.) production is constrained by fungal, bacterial and physiological disorders that reduce yield, fruit quality and market value. The conventional diagnosis is mainly based on visual inspection which is labour intensive and subjective. Moreover, the farmers may not be able to identify subtle or early symptoms. Artificial intelligence (AI), machine learning (ML) and deep learning (DL) offer non-destructive alternatives by learning disease-related patterns from images and other sensor data. This review consolidates current work on mango leaf disease detection, tracing the progression from handcrafted colour, texture and shape descriptors to convolutional neural networks (CNNs), transfer learning, hybrid classifiers, vision transformers and explainable AI. A systematic narrative search of five databases was conducted to identify studies published until April 2026 that used Boolean terms related to mango leaf disease and artificial intelligence; studies were screened and synthesised due to heterogeneity in dataset, disease classes, models, and validation methods. Reported studies show strong experimental performance for CNN-SVM systems, ResNet and MobileNet variants, severity classifiers and CNN-transformer models. However, headline accuracies frequently reflect curated datasets and do not necessarily demonstrate reliability in orchards, where illumination, background clutter, leaf occlusion, cultivar differences, mixed disorders and disease stage can alter symptom appearance. Beyond just accuracy, review emphasize the importance of quality datasets, independent validation, efficient reporting and deployment. It also encourages mobile and edge AI, severity assessment and integration with precision agriculture, as well as thermal and multimodal imaging. On-going research should be directed towards geographically diverse data, annotations from experts, temporal and multimodal learning, self-supervised representation learning, recurrent model updates, explaining predictions and farmer-focused decision support. Sustainable mango cultivation can only be realized by having AI tools that are accurate, interpretable, and accessible, which will involve interdisciplinary validation and incorporation of such tools with plant pathology and extension.