COMPLX-YOLO: A Novel YOLOv11-Based Framework for Mosquito Object Detection in Complex Backgrounds
Year: 2026
Type: conference-paper
Abstract: Mosquitoes are important vectors of infectious diseases, and accurate mosquito detection is essential for mosquito surveillance and vector-borne disease prevention. However, mosquito detection in field environments remains challenging due to the scarcity of public mosquito detection datasets, complex background interference, and the presence of insects with morphological characteristics similar to mosquitoes. To address these issues, a Xinjiang mosquito object detection dataset containing 1,100 images and 35,690 annotated mosquito targets was constructed through field sampling in the Ili and Altay regions of Xinjiang, China. Furthermore, a novel mosquito detection framework, termed COMPLX-YOLO, was proposed based on YOLOv11. The model introduces the RFCBAM module to enhance fine-grained feature representation and suppress background interference, employs the C3k2-CFBlock module to strengthen feature extraction capability, and incorporates PSConv and MANet-GCConv into the neck network to improve multi-scale feature fusion. Experimental results show that COMPLX-YOLO improves the mAP50 from 86.5% to 89.7%, achieving a 3.2 percentage point gain over the baseline YOLOv11 model. In addition, the proposed method demonstrates strong robustness and reliable detection performance under complex background conditions, providing an effective solution for intelligent mosquito monitoring.
Authors: Jinfa Zhong, Zhenguo Gao, Li Zhang, Gang Zhou
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202216358