Speaker
Description
Mosquito-borne diseases pose a major global health burden. Effective vector control depends on understanding how mosquitoes interact with their environment, making behavioral research essential. Object detection models can be trained to distinguish mosquitoes from complex backgrounds, offering high-throughput solutions for behavioral tracking. Here, we present YOLito (YOLO-Mosquito), an open-source detector based on YOLOv11, fine-tuned specifically for mosquito detection in both laboratory and real-world settings. Trained on over 100,000 annotated images from public datasets and research labs worldwide, YOLito detects multiple mosquito species under diverse conditions and achieves high performance on an unseen test dataset. In addition, we release a ready-to-use analysis toolkit that enables quantification of trajectory coordinates, visit frequency, duration, and distance. We also demonstrate its capabilities in a case study involving host-seeking female mosquitoes. YOLito provides a fast, accurate, and cost-effective tool for advancing behavior-based mosquito research and can improve vector control strategies.
