BEEHIVE: A dataset of Apis mellifera images to empower honeybee monitoring researchMendeley Data

Abstract

This data article describes the collection process of two sub-datasets comprehending images of Apis mellifera captured inside a commercial beehive (“Frame” sub-dataset, 2057 images) and at the bottom of it (“Bottom” sub-dataset, 1494 images). The data was collected in spring of 2023 (April–May) for the “Frame” sub-dataset, in September 2023 for the “Bottom” sub-dataset. Acquisitions were carried out using an instrumented beehive developed for the purpose of monitoring the colony's health status during long periods of time. The color cameras used were equipped with different lenses accordingly (liquid lenses for the internal one, standard lens of 8 mm focal length) and actuated by an embedded board, alongside red LED strips to illuminate the inside of the beehive. Images captured by the internal camera were mostly out-of-focus, thus a filtering procedure based on the adoption of focus measure operators was developed to keep only the in-focus ones. All images were manually labelled by experts using 2-class bounding boxes annotations representing full visible bees (class “bee”) and blurred or occluded bees according to the sub-dataset (“blurred_bee” or “occluded_bee” class). Annotations are provided in YOLO v8 format. The dataset can be useful for entomology research empowered by computer vision, especially for counting tasks, behavior monitoring, and pest management, since a few occurrences of Varroa destructor mites could be present in the “Frame” sub-dataset.

Keywords

• Entomology
• Precision agriculture
• Computer vision
• Object detection
• YOLO