AgriScan: Next.js powered cross-platform solution for automated plant disease diagnosis and crop health management

Abstract

Abstract Plant diseases present a formidable challenge to the agricultural sector worldwide, leading to significant losses, with the US experiencing annual losses amounting to one-third of crop production. Diagnosis of crop diseases through optical observation of leaf symptoms is particularly daunting for farmers with limited resources. Therefore, there is an urgent need for enhanced detection, monitoring, and prediction methods to mitigate agricultural losses effectively. Harnessing the power of computer vision and deep learning, this paper introduces a cross-platform system designed to automate plant leaf disease diagnosis. The system employs convolutional neural networks to classify 46 disease categories, trained on a dataset comprising 96,206 images of healthy and infected plant leaves. The user interface, accessible across multiple platforms including Android, iOS, Windows, and Linux, allows farmers to capture photos of infected leaves and receive real-time disease classification along with confidence percentages. By empowering farmers to maintain crop health and prevent the application of incorrect fertilizers, the system aims to optimize crop productivity. Performance evaluation includes metrics such as classification accuracy and processing time, with the model achieving an impressive overall accuracy of 93.45% across 46 common disease classes spanning 16 crop species.

Plant phenotyping relevance

葉画像から植物病害状態を推定するコンピュータビジョン手法と、そのクロスプラットフォーム実装・性能評価が中心であり、植物フェノタイピング手法として適格。

Code and data availability

The paper's plant-phenotyping input is a public Kaggle image dataset (Plant Diseases Training Dataset, ~96,206 leaf images across 16 crops/46+ disease classes) explicitly cited as dataset [11] and confirmed in the Availability of data and materials statement. No author analysis code, trained model checkpoints, or paper