Image-Based Detection of Plant Leaf Diseases Using Convolutional Neural Networks

Abstract

This study employs CNN models, specifically pretrained VGG-16 and VGG-19, to classify plant leaf diseases using transfer learning. For this, first the set of training plant leaf images used for the classification purpose has been preprocessed. Next, pre-trained VGG-16 and VGG-19 models are applied to the training plant leaf images using the steps of transfer learning. After training, the models are validated. Then, the models are used to classify the plant leaf diseases. Also, in the transfer learning process, fine-tuning (FT) is applied to retrain selected layers of the pre-trained CNN models (i.e., VGG 16 and VGG 19) to enhance performance. Thus, four models (Model 1: VGG 16 (without FT), Model 2: VGG 19 (without FT), Model 3: VGG 16 (with FT), and Model 4: VGG 19 (with FT)) have been created. In this work, the tomato plant leaves dataset has been considered. All the models have been used to classify the leaves into five categories: bacterial spot, early blight, late blight, leaf mold, and healthy. The models have been analyzed by comparing the classification accuracies produced by them.

Plant phenotyping relevance

植物葉の画像から病害状態を分類するCNN手法の開発・比較・検証が中心であり、植物表現型（病徴・病害状態）の画像ベース推定に該当する。

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.