Detection of Varroa Mites from Bee Images Using YOLO Architecture

Varroa destructor is considered one of the most detrimental parasitic species, as it weakens honeybees and transmits several deadly viruses that can lead to Colony Collapse Disorder (CCD). Therefore, early detection of Varroa mites in bee colonies is crucial, allowing beekeepers to take timely action to prevent their spread and minimize damage. YOLO (You Only Look Once) is a family of state-of-the-art object detection models, renowned for their high performance across various challenging benchmark datasets. However, a systematic assessment of these models for detecting Varroa mites from images is still lacking. This study aims to comprehensively evaluate several YOLO models for detecting Varroa mites. Six models were selected and evaluated on the publicly available VarroaDataset and a custom-built dataset. Experimental results demonstrate that YOLO architectures achieve high detection accuracy for Varroa destructor with low inference times, highlighting their strong potential for use in real-time automated bee monitoring systems.