A Compact Deep Learning Approach Integrating Depthwise Convolutions and Spatial Attention for Plant Disease Classification

Agriculture is a major sector that provides food to the growing worldpopulation, so improving crop yields from the disease is essential. Bac-terial, fungi and viral plant diseases significantly challenge high cropproductivity. Traditional methods of plant disease detection are time-consuming and require expert knowledge. Therefore, to ensure optimalcrop health and yield, advanced, efficient, and accurate methods areneeded to detect and classify plant diseases early and accurately. Thisreview article explores the recent advancements in vision-based machinelearning (ML) and deep learning (DL) techniques for detecting plantleaf diseases. Specifically, the objective is to provide a detailed overviewof the current state-of-the-art methodologies, assess their effectiveness,and identify the role of critical datasets in facilitating these advance-ments. Additionally, the article highlights the potential of integratingemerging technologies, such as vision based models, to enhance diseasedetection systems’ capabilities. The review systematically examines var-ious ML and DL techniques for plant leaf disease detection, focusingon their application and performance. Techniques such as Convolu-tional Neural Networks (CNNs), which have shown high accuracy in disease identification, are discussed in depth. The importance of open-source databases, particularly PlantVillage, in training and validatingthese models is emphasized. The methodology also discusses integrat-ing advanced technologies and developing hybrid models that combinevision-based and language-based approaches to improve detection accu-racy and robustness. The analysis highlights the significant progressin applying ML and DL to plant leaf disease detection, showcasingseveral successful implementations and their outcomes. The reviewidentifies emerging trends that promise to enhance disease detectionsystems. It also critically assesses the strengths and limitations ofcurrent techniques, offering insights into potential improvements andfuture research directions. This comprehensive analysis highlights thecritical role of ML and DL in advancing plant disease detection, aim-ing to provide sustainable and scalable solutions for the agriculturalsector. The findings have significant implications for improving cropyields and ensuring food security for the growing global population.