Deep Learning-Based Crop Disease Detection for Precision Agriculture - A Survey

Abstract

Crop diseases continue to pose a serious challenge to global agricultural productivity, leading to substantial yield losses, economic instability, and threats to food security. Conventional crop disease detection methods rely heavily on manual visual inspection by farmers or experts, which is time-consuming, subjective, and impractical for large-scale and continuous monitoring. In response to these limitations, recent advancements in precision agriculture have encouraged the adoption of intelligent and automated techniques for crop health assessment. This review paper critically examines a dissertation that presents a deep learning-based framework for crop disease detection using convolutional neural networks (CNNs). The reviewed study employs image-based analysis of crop leaf images and formulates the problem as a binary classification task, distinguishing between healthy and diseased crops. The proposed system integrates image preprocessing techniques with hierarchical feature extraction through CNN architectures, eliminating the need for handcrafted features. Model performance is evaluated using standard classification metrics, including accuracy, precision, recall, F1-score, and confusion matrix analysis. Experimental findings demonstrate an overall classification accuracy of 93.75 percent, accompanied by balanced precision and recall values across both classes, indicating strong generalization and reliable disease detection capability. This review synthesizes the methodology, experimental outcomes, and significance of the study, while also highlighting existing limitations and potential directions for future research in intelligent precision agriculture systems

Plant phenotyping relevance

植物葉画像から健全・罹病状態を推定する画像ベース手法を中心に扱うレビューであり、植物病害状態のフェノタイピング手法に該当する。

Code and data availability

The article is a survey/review of a dissertation on CNN-based crop disease detection. It describes a 2000-image dataset and a compact CNN, but provides no public dataset link, no code availability statement, no repository, and no supplement. All URLs in the text are bibliographic references to cited prior work, not the

No evidence-backed public reproduction asset is currently recorded.