摘要
Abstract
To address the issues of large parameter quantity and high computational complexity in existing tomato disease detection models,this paper proposes a lightweight detection algorithm,GEC-YOLOv11.This model is based on YOLOv11n and replaces the C3K2 module in the backbone network with the C3K2-GhostModule to reduce the parameter quantity and computational complexity.It introduces C2SPA-EDFFD to enhance the local feature extraction capability while maintaining the original global feature extraction abili-ty.Additionally,the neck network is optimized based on CCFM to improve feature fusion efficiency and significantly reduce the model size.Experimental results show that compared with YOLOv11n,GEC-YOLOv11 reduces the parameter quantity,FLOPs,and model weights by 35.1%,23.8%,and 34.0%respectively,with an mAP of 93.14%and a recall rate increased to 86.58%.Compared with models such as YOLOv8n and YOLOv10n,GEC-YOLOv11 demonstrates superior overall performance.Visualization results indicate that this model performs well in detecting small targets and in dense scenes,with a significantly reduced missed detection rate.The improved mod-el in this study can meet the requirements of tomato leaf disease detection tasks in agricultural scenarios in terms of accuracy and light-weight,providing a solution for the lightweight detection of tomato diseases.关键词
YOLOv11/番茄病虫害/轻量化/深度学习Key words
YOLOv11/Tomato disease and pest/Lightweight/Deep learning分类
信息技术与安全科学
