A Novel Approach for Plant Disease Detection by Using Zero Short Learning

Abstract

One of the most significant issues in identifying plant diseases is agriculture today, as its implications are tremendous for crop health and yield. This paper addresses the challenge of detecting visible and invisible illness classes by combining the YOLO-NAS model with the Zero-Shot Learning (ZSL) framework. The approach leverages semantic attributes and object detection capabilities to classify diseases without extensive retraining, thus allowing scalability and adaptability. The success of the suggested strategy is shown by the major performance indicators, which include accuracy, recall, mean average Precision (mAP), and F1-score. As stated by experimental findings, the ZSL model outperforms traditional object detection algorithms, with a mAP of 0.92, a recall of 0.91, and an accuracy of 0.92. at a threshold of 0.5. A comparative analysis with the YOLOv5 and YOLOv8 models highlights the better balance of accuracy, recall, and efficiency that the proposed ZSL-YOLO-NAS model can achieve. This study thus forms a significant step in automatically detecting plant diseases and provides a strong and dependable agricultural application solution.

DOI - https://doi.org/10.65525/SVUP.9788199778009.2026.172-194