VGDNet : An Attention-Based Weighted Fusion Architecture for Automated Tomato Leaf Disease Detection with Explainable AI
Year: 2026
Type: conference-paper
Abstract: Tomato leaf diseases substantially affect global agricultural productivity, highlighting the necessity for automated and precise detection technologies to facilitate sustainable crop management. This paper introduces VGDNet , a novel attention-driven weighted fusion architecture for the robust classification of tomato leaf diseases, named after the combination of VGG-16 and DenseNet-121. VGDNet employs a specific attention mechanism that adaptively highlights pathologically significant characteristics, hence improving representational effectiveness in contrast to conventional methods. The model was assessed using a comprehensive hybrid dataset consisting of eight categories: seven principal disease classes and healthy leaves. Experimental findings indicate that VGDNet attains 99% in accuracy, precision, recall, and F1-score, surpassing the performance of proven designs like VGG-16 and DenseNet-121 separately. Moreover, the amalgamation of Grad-CAM and Grad-CAM++ visualization methodologies improves interpretability by distinctly localizing disease-specific areas. The findings underscore the efficacy of VGDNet in automated plant disease diagnosis, providing enhanced classification performance and more explainability, hence indicating potential for practical applications in precision agriculture.
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: hybrid
OpenAlex ID: W7204481820