Comparative Investigation of Deep Convolutional Networks in Detection of Plant Diseases

Abstract

Abstract: Preserving plant health and early detection of diseases are crucial in modern agriculture. Artificial intelligence techniques, particularly deep learning networks, are employed for this purpose. In this study, disease recognition was conducted using leaf images from various plant species. The study encompassed important agricultural products such as apples, strawberries, grapes, corn, peppers, and potatoes among the plant species considered. Among the deep learning networks, popular architectures like AlexNet, Vgg16, MobileNetV2, and Inception were compared. The Inception V3 model achieved the highest success rate of 92%, followed by the AlexNet architecture with a success rate of 91%. Among these networks, the InceptionV3 model yielded the best results. The InceptionV3 model effectively learned from plant leaf images and accurately distinguished between diseased and healthy leaves. These findings demonstrate that AI-based systems can be efficiently utilized for disease recognition and prevention in the agriculture sector. In this study, the performance of the InceptionV3 model in disease recognition on plant leaves was analyzed in detail, emphasizing the role of deep learning networks in agricultural applications.

Plant phenotyping relevance

植物葉画像から健全・罹病状態を推定する深層学習手法を比較・評価しており、植物病害表現型の取得・分類が研究の中心です。

Code and data availability

The paper's plant-disease classification experiments were performed on the public New Plant Diseases Dataset (Kaggle), which the authors explicitly state is openly accessible via a Kaggle URL. This is a paper-specific, public, actionable phenotype image dataset. No author analysis code or trained models are reported as