YOLO-based Bee-Hornet Real-Time Notification

Daily inspection of bee colony status and monitoring for hornet attacks are important tasks in apiary management. Traditionally, these tasks require beekeepers to regularly check the apiaries, which is a labor-intensive and knowledge-intensive operation. With the advancement of the Internet of Things (IoT) and machine learning, we can now use modern technology to improve the efficiency of these operations. This paper designs and implements a bee and hornet recognition and real-time notification system to reduce labor costs and improve apiary management. By collecting online and onsite photos and videos of bees/hornets, we trained a YOLO v7 model. The trained model is then downloaded to a Raspberry Pi device, which uses a web camera to capture onsite photos and videos for system recognition and notifies the beekeeper. To avoid excessive or uncertain notifications that may disrupt the beekeeper's work and cause anxiety, we designed an appropriate algorithm to adjust the notification threshold and frequency. After testing with actual data, the precision and recall rates of our bee and hornet recognition system and the notification time meet the requirements.