Artificial Intelligence in Pathogen Based Plant Disease Detection: Advances, Trends, and Challenges

Abstract

Plant disease outbreaks caused by fungi, bacteria,and viruses continue to threaten crop yields, food security, andthe long-term health of agricultural systems across the globe.Detecting these diseases early, accurately, and at scale remains adifficult task, especially as climate patterns shift and food supplychains stretch across continents. Conventional diagnostic methodssuch as visual inspection, lab-based tests, and molecular assayshave long been the standard, yet they are often slow, expensive,and tied to controlled laboratory settings. In recent years,advances in artificial intelligence (AI) have begun to transformthis landscape. Modern approaches powered by convolutionalneural networks (CNNs), transformers, and multimodal AI nowtake on the challenge of interpreting complex agricultural data,whether from images, sensors, or text. This review explores thecurrent ecosystem of AI-driven techniques for detecting plantdiseases caused by pathogens, comparing how different methodsperform and where each one fits best. Special focus is givento the shift from traditional image-based deep learning towardnewer frontiers such as explainable AI (XAI), graph neuralnetworks (GNNs), few-shot and foundation models, and fullymultimodal systems. We structure our comparison not aroundaccuracy rankings alone, but around what each paradigm canand cannot do: the conditions under which it was tested, what itrequires to operate, what it reports about its own decisions, andhow close it has come to real agricultural deployment. Along theway, the review identifies persistent challenges including datasetover-reliance on controlled benchmarks, poor domain transfer toreal-world conditions, constraints in real-time deployment, andthe absence of standardised explainability evaluation. Drawingfrom the latest research, it outlines the major gaps that stillremain and offers practical directions for future work.