A Review on Deep Learning Approaches for Plant Leaf Disease Detection and Pesticide Recommendation

Abstract

Crop sustainability is an urgent issue in the world, and plant diseases are a significant reason for crop yield and food security restrictions. Conventional manual disease detection procedures are time-consuming and frequently inaccurate. Deep learning, particularly Convolutional Neural Networks (CNNs), has made automated and accurate plant disease diagnosis from leaf images possible in recent years. This review overviews current deep learning methods for the detection of plant leaf diseases and delves into models with pesticide recommendation systems. The review classifies studies into CNN-based models, classical machine learning methods, hybrid models, and image processing approaches. It also identifies systems with region-specific pesticide recommendations. Issues like data shortage, generalization of models, and compliance with regulations are addressed, in addition to directions for the future like mobile deployment, transfer learning, and IoT integration. This thorough review is designed to direct future research toward smart, scalable agriculture solutions.

Plant phenotyping relevance

植物葉画像から病害状態を推定する深層学習手法を中心にレビューしており、植物フェノタイピング手法レビューに該当する。農薬推薦も扱うが、葉病害検出の方法論が明示的な中心である。

Code and data availability

This is a review paper with no original phenotyping measurements, datasets, code, models, or supplements; it only cites third-party studies and the generic PlantVillage dataset without any authors' public asset URLs.

No evidence-backed public reproduction asset is currently recorded.