Designing an AI-Based Greenhouse Plant Monitoring System to Detect and Classify Plant Diseases from Leaf Images

Abstract

Plant diseases can significantly hinder food crop production, leading to substantial economic losses and posing a threat to global food security. Machine learning, particularly deep learning, plays a crucial role in object detection and classification. In this study, we present an AI-based plant monitoring system for detecting and classifying plant diseases using visual images.Our deep learning models are trained on plant images obtained from natural environments. Manual detection and classification are both challenging and labor-intensive, making accurate and timely diagnoses from an automatic system highly beneficial for treating plant diseases. Traditionally, plant disease detection using deep learning has relied on images taken in controlled environments, which do not support in-situ detection for remote monitoring. The Plantdoc dataset, a popular resource consisting of plant images from actual field conditions, is used in our study.We employ the YOLOv5 algorithm from the field of computer vision to the Plantdoc dataset, achieving results that surpass previous work on the same dataset. This success is attributed to our selected model and data augmentation techniques. Our model can classify and detect various diseased and healthy leaf classes with a mean Average Precision (mAP) of 92%. This capability enables farmers and researchers to remotely monitor plant health and diagnose plant diseases, thereby saving time, reducing costs, and minimizing crop loss.

Plant phenotyping relevance

葉画像から植物の病害状態を検出・分類するAI手法を開発・評価しており、植物表現型の取得が研究の中心です。

Code and data availability

The paper uses the public PlantDoc dataset and YOLOv5 via Roboflow/Google Colab, but provides no authors' public code, trained model checkpoints, or paper-specific data deposit. PlantDoc is cited prior work, not a paper-specific asset.

No evidence-backed public reproduction asset is currently recorded.