Research library

Visual recognition of honeybee behavior patterns at the hive entrance

This study presents a method for automatically recognizing honeybee behavior patterns at the hive entrance. Using YOLOv8 models for detection and segmentation, the approach analyzes bee location, direction, path trajectory, and movement speed on the hive landing board. The system detects multiple activities, including foraging, fanning, washboarding, and defense, achieving a mean detection accuracy of 98% and up to 36 fps. Key contributions include a dataset with 7200 frames from eight beehives, behavior-class tracks, and a comparative evaluation of object detection and tracking algorithms tailored for bee detection and behavior recognition.

Download PDF

Publication details

Authors
Tomyslav Sledevič, Artūras Serackis, Dalius Matuzevičius, Darius Plonis, Gabriela Vdoviak
Organizations
🇱🇹 Vilnius Gediminas Technical University
Year
2025
Type
Journal

Relevancy to Gratheon

This paper is directly actionable for Gratheon's entrance-camera and gate-tracker concepts. It moves beyond simple in/out counting by linking object detection, segmentation, trajectories, speed, direction, and density maps to recognizable hive-entrance behaviors such as foraging, fanning, washboarding, and defense. The open dataset is also valuable for benchmarking Gratheon models on real entrance footage and for designing behavior alerts in the web app without opening the hive. The reported 98% mean detection accuracy and up to 36 fps provide useful performance targets for real-time monitoring deployments.