Deep Learning-Based Crop Disease Recognition System for Smart Agriculture

Abstract

With the rapid advancement of artificial intelligence (AI) and computer vision, intelligent agricultural systems have become a crucial component of smart farming. Among them, automatic crop disease recognition plays a vital role in ensuring agricultural productivity and food security. This study proposes an AI‑based crop disease recognition system that integrates deep learning, image processing, and edge computing. A large‑scale dataset of crop disease images was constructed, and transfer learning was employed to enhance model generalization. A convolutional neural network (CNN) was optimized by incorporating attention mechanisms and multi‑scale feature fusion to improve accuracy. Experiments show an average accuracy of 97.8% on the PlantVillage dataset [9] and stable performance under real‑field lighting variations. A lightweight deployment framework based on TensorFlow Lite enables real‑time disease detection on mobile and embedded platforms. The system provides a feasible, efficient AI‑driven solution for precision agriculture and contributes to the digital transformation of modern farming.

Plant phenotyping relevance

植物病害画像から病害状態を推定する深層学習システムの開発・検証が中心であり、植物フェノタイプ測定に該当する。

Code and data availability

The preprint describes a crop disease recognition system using PlantVillage data and a 62,000-image dataset, but provides no data availability statement, no public repository, no code/model release, and no author-provided URLs. The only dataset mentioned (PlantVillage) is cited prior work, not a paper-specific asset.

No evidence-backed public reproduction asset is currently recorded.