Integration of YOLO11 into scale insect detection and quantification

The You Only Look Once (YOLO) machine learning method for object detection and quantification is increasingly used for insect detection and counting. However, insects in the superfamily Coccoidea remain poorly studied in this context, despite representing some of the most invasive and difficult to detect pests on the globe. Rapid detection and quantification of insects from the families Coccidae and Diaspididae could be transformational for agricultural inspectors and researchers working with scale insect pests. Here, we use the YOLO11 model by Ultralytics to identify and quantify adult females of three different scale insect species found on palms, Fiorinia phantasma, Chrysomphalus aonidum, and Aonidiella orientalis. We used a combination of publicly available images and images taken from the field, which were expanded using an augmentation pipeline. We compared the speed and accuracy of scale quantification by the trained YOLO model to that of a human counter. We also evaluated practical applications by integrating new species into the model’s detection training, compared to training species detection from scratch. We hope to build an image database that is accessible and adaptable for new research on scale insects.