YOLOv11-GPC: A Lightweight Detection Model for Maize Diseases and Pests with Edge-Device Deployment on Jetson Orin Nano

Abstract

Maize faces a wide variety of pests and diseases, and its growing environment is complex, which makes detecting these problems quite challenging. Existing object detection algorithms usually require a lot of computing power, but edge computing devices are limited in terms of processing resources and memory, making it hard to meet the real-time and efficient detection needs for maize pests and diseases. To tackle this, our study designed an edge computing detection system for maize pests and diseases based on an improved YOLOv11 model. First, we integrated PConv into the Bottleneck module to build the C3k2-PConv module, replacing the C3k2 module in YOLOv11’s neck network, which reduces model parameters and improves detection efficiency for maize pests and diseases. Next, we introduced the GhostConv module to replace standard convolutions in the backbone network, which lowers computational complexity while maintaining high detection accuracy. Additionally, we designed a C2PSLCA module based on linear self-attention to replace the C2PSA module in YOLOv11, enhancing the model’s ability to extract features of maize pests and diseases in complex environments. The principal role of C2PSLCA is feature representation rather than a large reduction in model complexity. The results show that the improved YOLOv11 model can detect maize pests and diseases efficiently and accurately, achieving a precision of 87.65%, recall of 81.19%, and mAP@0.5 of 87.08%, which are improvements over the YOLOv11n baseline of 0.78, 1.96, and 2.28 percentage points, respectively. The performance improvements reported in this study are relative to the YOLOv11n baseline only. Meanwhile, model weight, parameters, and FLOPs were reduced by 14.02%, 14.89%, and 19.40%, boosting detection efficiency and deployment feasibility. Finally, we deployed the improved YOLOv11-GPC model on the Jetson Orin Nano platform and used PySide6 to develop a system interface, providing technical support for maize health management and pest and disease control.