Quantifying Disease Severity in Groundnut Plant through UNet3+ and Vision Transformers

Accurately assessing disease severity in groundnut plants is a critical challenge in agricultural technology and plant pathology.This project addresses the need for early and precise detection by proposing an integrated approach combining Vision Transformers (ViTs) and UNet3+. ViTs are utilized to classify groundnut leaf images based on disease severity, effectively capturing complex patterns for accurate categorization. Subsequently, the UNet3+ model segments the diseased regions within the leaves, offering clear visual insights into affected areas. The severity ratio is then computed by measuring the proportion of diseased pixels relative to the total leaf area. This value is further scaled using the Horsfall-Barratt scale to ensure standardized severity grading. By merging classification and segmentation, this approach enhances the precision of disease assessment and provides a detailed understanding of disease spread with 94 percent accuracy. The system enables timely and informed decision-making for groundnut crop management, ultimately contributing to improved yield and sustainable agricultural practices.