Deep learning techniques for plant disease detection and classification: A comprehesive review

Abstract

India's expanding population and rising food needs depend heavily on agriculture, therefore raising agricultural yields is essential. Bacterial, fungal, and viral plant diseases are major contributors to decreased agricultural productivity. Beyond labour-intensive classical methods, machine learning (ML) and deep learning (DL) techniques provide potential alternatives for precise and effective plant disease identification. Through an analysis of multiple published research articles, this study thoroughly examines the use of DL approaches for plant disease detection and control in agriculture. In addition to a thorough analysis of publically accessible datasets and well-known DL architectures, the review concentrates on disease classification, detection, and segmentation. DL-human accuracy comparisons, imaging sensors, model generalisation, illness severity estimate, and dataset needs are important factors. The results demonstrate how DL, when trained on sizable, high-quality datasets, can accurately identify plant diseases in their early stages. Challenges like data collecting, model scalability, and incorporation into agricultural operations are also identified in this review. By filling in these gaps, DL-based technologies can improve precision farming, helping farmers manage diseases and promoting food security and sustainable agriculture. This thorough evaluation provides insightful information and recommendations for improving plant disease detection methods in contemporary agriculture.