Review of a Decentralized Ensembled Deep Learning Model for Early Detection and Control of Tomato Plant Diseases

Abstract

This study proposes a decentralized ensemble deep learning system for the early detection and management of tomato plant diseases, combining the capabilities of Convolutional Neural Networks (CNN), Vision Transformers (ViT), and U-Net architectures. The integrated model is designed to overcome challenges related to the timely and precise identification of plant diseases by harnessing the strengths of each component: the CNN extracts crucial local features from leaf images, the ViT captures broader relationships and complex patterns across the image, and the U-Net performs detailed segmentation to accurately locate affected regions. The decentralized nature of the system is supported by blockchain-based smart contracts, which facilitate secure, transparent, and efficient data exchange among stakeholders. By operating across distributed nodes, the ensemble approach enables real-time disease recognition at the edge, enhancing scalability and robustness without dependence on a central server. Experimental results indicate that this model delivers superior accuracy, speed, and reliability compared to traditional methods, offering a promising pathway for proactive and collaborative tomato disease management. The architecture not only ensures data integrity but also encourages data sharing among farmers, agronomists, and researchers, advancing the development of intelligent and sustainable agricultural solutions.