Detection and Classification of Plant Leaf Diseases

Abstract

Plant diseases pose a serious threat to global food production, leading to yield losses, increased production costs, and environmental damage. Plant pests and diseases have widespread negative impacts on economic, ecological, environmental, and human health, and calculating the true cost of these damages is extremely difficult. Traditional diagnostic methods also present significant limitations in terms of time and cost. This study adopts an image processing-based approach to this problem. Using the Crop Disease Detection dataset, plant disease detection was performed using CNN, DNN, K-NN, SVM, XGBoost, and Random Forest algorithms, employing both deep learning and machine learning methods. The study demonstrates that CNN architectures designed from scratch, without resorting to pre-trained models such as ResNet and MobileNet, can also exhibit high performance. The highest accuracy rate was obtained with the CNN model at 94.08%. In machine learning models, grid search was used for hyperparameter optimization, and the best results were achieved through this method.

Plant phenotyping relevance

植物葉の画像から病害状態を推定する画像処理・機械学習手法を比較評価しており、病害フェノタイピング手法が研究の中心である。

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.