Salinity-resilient crop health monitoring: automated disease detection in luffa Aegyptiaca leaves using Vision Transformer & CNN

Abstract

In coastal areas where agriculture is often constrained by salinity, cultivating specific crops like Luffa Aegyptiaca (sponge gourd) becomes crucial for local sustenance. Identifying these diseases is challenging and time-consuming when no domain specialists are present accurately, and the information needs to be more consistent. Effective disease detection and management play a pivotal role in ensuring the viability of these limited yet vital crops, impacting crop yield, fertilization strategies, and overall food security for coastal communities. This groundbreaking study focuses on detecting and classifying leaf diseases within Luffa Aegyptiaca leaves, prevalent crops in coastal regions. Leveraging the cutting-edge capabilities of Convolutional Neural Networks (CNN) and Vision Transformer algorithms, our research achieves unparalleled accuracy. The CNN algorithm boasts an impressive accuracy of 98.32%, while the Vision Transformer algorithm surpasses expectations with an exceptional accuracy of 99.85%. Notably, this study utilizes an original dataset, a unique contribution to the field given the absence of publicly available datasets or prior research specific to Luffa Aegyptiaca. Beyond mere accuracy metrics, our findings illuminate profound insights into the nuanced landscape of leaf disease detection and classification, affirming the remarkable efficacy of both CNN and Vision Transformer algorithms. In conclusion, this research advances our understanding of plant pathology, and underscores the unparalleled potential of state-ofthe-art machine-learning techniques in agricultural research.