A High-Accuracy Automated Plant Disease Classification System Using CNN Architectures and Web Deployment

Abstract

Agricultural productivity faces significant challenges due to plant diseases caused by microscopic pathogens that are difficult to detect during early developmental stages. Traditional disease detection methods rely heavily on manual inspection by agricultural experts, which is time-consuming, expensive, and prone to human error. This research presents an automated plant disease detection system that leverages deep learning and transfer learning to identify and classify plant diseases with high accuracy. The proposed framework utilises the comprehensive Plant Village dataset, which contains 54,309 images spanning 14 crop species and 38 disease classes. Multiple convolutional neural network architectures, including AlexNet, VGG16, InceptionV3, and MobileNet, were implemented and evaluated using colour, grayscale, and segmented image representations. The models were trained using an 80-20 split between training and test data to ensure robust performance evaluation. Performance metrics, including accuracy, precision, recall, and F1-score, were computed to assess model effectiveness. Among all architectures, AlexNet achieved the highest performance, with 99.56% accuracy, 0.9953 precision, 0.9971 recall, and 0.9961 F1-score. The system is deployed as a web application using HTML, CSS, JavaScript, Keras, and Python, hosted on Heroku. This automated solution enables farmers to detect plant diseases at early stages with minimal cost, facilitating timely intervention and improved crop yield management.

Plant phenotyping relevance

植物画像から病害状態を分類するCNN手法を開発・比較し、性能評価とWeb実装まで行っており、植物フェノタイピング手法が中心である。

Code and data availability

The paper uses the public PlantVillage dataset (54,309 images) and CNN models, but provides no authors' code, trained model checkpoints, or dataset deposit URL. No explicit availability/deposit language or public repository link for the authors' analysis or web application appears in the supplied blocks, so no paper-­‐

No evidence-backed public reproduction asset is currently recorded.