A Deep Learning-Based Decision Support Tool for Plant-Parasitic Nematode Management.

Area of Science:

• Agricultural Science
• Computer Science
• Nematology

Background:

• Plant-parasitic nematodes (PPN), particularly root-knot nematodes (RKN), cause significant crop yield losses and economic damage globally.
• Current methods for identifying and quantifying PPN populations are labor-intensive and time-consuming, hindering effective management.

Purpose of the Study:

• To develop a state-of-the-art deep learning model for detecting and estimating plant-parasitic nematode populations.
• To create a user-friendly decision support tool (NemDST) for farmers to manage nematode infestations.

Main Methods:

• Utilized the YOLOv5 deep learning model with pre-trained weights for detecting RKN juveniles and eggs.
• Integrated the YOLOv5 model into a web application to create the NemDST prototype.
• Evaluated model performance using precision, recall, F1-score, and mean Average Precision (mAP).

Main Results:

• The YOLOv5-640 model achieved high accuracy in detecting RKN eggs (precision=0.992, recall=0.959, F1-score=0.975, mAP=0.979).
• Detection inference time was rapid at 3.9 milliseconds, outperforming other methods.
• The NemDST system provides image input, population assessment, growth tracking, and control recommendations.

Conclusions:

• The NemDST tool offers a fast and reliable solution for assessing nematode populations.
• This technology has the potential to significantly reduce crop yield losses and enhance agricultural economic outcomes.
• Automated nematode detection and population estimation can revolutionize pest management strategies.