Hybrid CNN-Based Plant Disease Detection for Smart Agriculture: A Review

Abstract

Abstract Plant diseases are a major challenge in agriculture, leading to reduced crop yield and economic losses. Traditional methods of disease detection rely on manual inspection, which is time-consuming and less accurate. With the advancement of Deep Learning, Computer Vision, and Internet of Things, automated plant disease detection systems have been developed to improve accuracy and efficiency. Recent research shows that deep learning models, especially Convolutional Neural Networks (CNNs), are widely used for plant disease classification due to their ability to automatically extract features from images and achieve high accuracy [3], [34] . Transfer learning techniques using models such as ResNet and VGG further enhance performance, particularly when datasets are limited [9], [41] . In addition, lightweight models like MobileNet and ShuffleNet are being developed for deployment on mobile and IoT devices [2], [12] . Many studies use benchmark datasets such as PlantVillage, which provide high accuracy under controlled conditions. However, real-world applications face challenges such as varying lighting conditions, complex backgrounds, and limited dataset diversity [13], [15] . To address these issues, researchers are exploring advanced techniques such as data augmentation, object detection models like YOLO, and multimodal approaches that combine image data with environmental sensor data [5], [10] . IoT-based systems are also gaining importance as they enable real-time monitoring of crop conditions using sensors and smart devices [21], [26] . Furthermore, Explainable AI techniques such as Grad-CAM are being used to improve model transparency and help users understand the prediction results [25] . Despite significant progress, challenges remain in terms of model generalization, computational complexity, and real-time deployment. Future research should focus on developing lightweight, efficient, and scalable models, along with the use of large real-field datasets and edge computing technologies. Keywords Plant Disease Detection; Leaf Image Analysis; Convolutional Neural Network (CNN); Deep Learning; Transfer Learning; Computer Vision Internet of Things; Lightweight Models; Object Detection (YOLO); Multimodal Data Fusion; Explainable Artificial Intelligence (XAI)

Plant phenotyping relevance

植物病害を葉画像から推定するCNN・画像解析手法を体系的に扱うレビューであり、植物の病態を観測・推定するフェノタイピング手法が中心です。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.