AgriVision: A Vision Transformer-Based System for Plant Disease Detection and Advisory

Plant diseases pose a significant threat to global agricultural productivity, leading to substantial economic losses and food security challenges. Traditional disease detection methods rely heavily on manual inspection by experts, which is timeconsuming, labor-intensive, and often impractical for large-scale farming. Recent advancements in deep learning have enabled automated disease detection; however, most existing systems are limited to standalone classification models without real-world integration. This paper presents AgriVision, a Vision Transformer (ViT)based plant disease detection and advisory system. The proposed approach leverages the self-attention mechanism of transformers to capture global contextual features from plant leaf images, improving classification performance compared to conventional Convolutional Neural Networks (CNNs). In addition to disease prediction, the system integrates an officer validation module to enhance reliability and a marketplace module to provide farmers with access to recommended agricultural products. The model is trained and evaluated on multiple plant disease datasets using standard performance metrics such as accuracy, precision, recall, and F1-score. Experimental results show that the proposed model achieves a test accuracy of 94.21%, with a precision of 94.48%, recall of 94.21%, and F1-score of 94.01%, outperforming the baseline CNN model.. The overall system provides an end-to-end solution that bridges the gap between disease detection, diagnosis, and treatment, contributing to the development of scalable smart agriculture systems.