Key points are not available for this paper at this time.
Accurate and rapid identification of agricultural pests is essential for intelligent pest monitoring. However, existing pest detection models often suffer from high parameter counts and computational complexity, limiting their deployment on edge devices. To address these challenges, this paper proposes a lightweight agricultural pest detection model, YOLO-DCPG, based on YOLOv8n. First, a Dual Channel Pooling Gated Attention (DCPGAttention) module is designed, applying mean and standard deviation pooling to enhance global information capture. A lightweight backbone network, StarNet, is employed for feature extraction, while a feature fusion neck, Small-Neck, is introduced, based on an improved bidirectional feature pyramid network (a-BIFPN) and integrating the efficient GSConv module. This design reduces model parameters and computational cost while maintaining detection accuracy. Furthermore, a scale factor based on Inner-IoU is incorporated into the WIoU loss function, enabling more precise control over auxiliary bounding boxes and improving regression for small pest regions. Experimental results on the Pest24 dataset show that YOLO-DCPG achieves a precision of 80.1%, mAP@50 of 74%, and mAP@50~95 of 47.5%, representing improvements of 4.5%, 0.8%, and 0.9% over the baseline YOLOv8n, respectively. Meanwhile, the number of parameters, GFLOPs, and model size are reduced by 51.2%, 30.1%, and 46.7%, respectively. Finally, YOLO-DCPG is successfully deployed on Raspberry Pi 4B, achieving stable, real-time pest detection, demonstrating its effectiveness and practicality for edge agricultural applications.
Liu et al. (Thu,) studied this question.