EC-YOLO: Enhanced YOLOv10 for Agricultural Pest Detection

Agricultural pests pose a significant threat to agricultural production. There is an urgent need for efficient and accurate pest detection methods to guide rational pesticide application. Traditional pest detection relies on on-site inspections by experts, which are costly and inefficient. In recent years, deep learning technologies have made substantial progress in pest recognition but still struggle with detecting multiple types of pests in complex natural environments. To address these challenges, this paper introduces a new model based on YOLOv10, called ECYOLO, which makes the following key improvements:Firstly, this paper designed an Edge Information Perception Enhancement module to replace the traditional Cross Stage Partial Bottleneck with 2 Convolutions module. This module integrates multi-scale pooling branches and edge information extractor module, effectively capturing multi-scale features and edge information in images. Secondly, this paper designed the Cross-Stage Weighted Feature Pyramid Network module, which fuses feature maps of different scales and employs an adaptive weighting mechanism to enhance the effectiveness of feature fusion.Experimental results indicate that EC-YOLO outperforms the YOLOv10 model on the IP102 datasets, with an mAP50 improvement of 2.6%, reaching 65.8%, and a recall improvement of 2.1<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, reaching 63.3%. Additionally, EC-YOLO surpasses six other popular object detection models, demonstrating superior performance.