Deep Learning and Computer Vision for Banana Leaf Disease Diagnosis: A Systematic Survey and Conceptual Design Blueprint for Edge-Deployable Systems.
Year: 2026
Type: book
Abstract: Bananas (Musa spp.) are critical agricultural crops for global food security. Foliar diseases like Sigatoka, Cordana, and Pestalotiopsis significantly reduce crop yields, threatening food security and farmers' livelihoods. While deep learning (DL) has shown high accuracy under controlled laboratory settings, real-world deployment is hindered by domain shift, resource constraints on edge devices, and a lack of explainability. This paper presents a systematic literature survey and conceptual design blueprint for deep learning-based diagnostics in banana cultivation. We review the core computer-vision pipeline, comparing preprocessing, segmentation, and deep classification backbones on the "Banana Leaf Spot Diseases (BananaLSD)" dataset, highlighting severity estimation and treatment recommendations. We systematically analyze cross-domain techniques from 37 general plant pathology studies (covering wheat, rice, corn, tomato, etc.), and synthesize these insights to propose a conceptual architecture blueprint, Edu-ViTNet: a quantized, edge-deployable, hybrid CNN-Vision Transformer framework featuring Grad-CAM explainability and federated learning. Critically, rather than claiming immediate empirical performance for this proposed model, we frame it as a theoretical design blueprint and propose a rigorous validation protocol to guide future implementation. This study serves as a comprehensive roadmap for researchers aiming to develop robust, field-deployable smart-farming applications..
Authors: Anju Markose, K. Baalaji K. Baalaji
Institution: Bharath University
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: gold
OpenAlex ID: W7204924914