10.60507/FK2/MCQXPP Behley, Jens Jens Behley https://orcid.org/0000-0001-6483-0319 University of Bonn Chong, Yue Linn Yue Linn Chong https://orcid.org/0000-0002-5851-953X University of Bonn bonndata 2026 Agricultural Sciences Behley, Jens Jens Behley University of Bonn 2026-09-15 2026-09-28 10.1101/2025.10.27.682286 10.60507/FK2/MCQXPP/IGO8PK 10.60507/FK2/MCQXPP/YLR70F 10.60507/FK2/MCQXPP/7PGPFK 7027636 12785485688 9633 application/zip application/zip text/markdown 1.0 Creative Commons Attribution 4.0 International License. Pollinator monitoring is essential for understanding and addressing the decline of insect species and their populations. While computer vision techniques provide new tools for monitoring pollinators, there is a need to develop well-annotated datasets to support artificial intelligence (AI) methods for pollinator detection and to assess their effectiveness. Creating such datasets is challenging because pollinators are small and difficult to detect among plants, even for human annotators. We propose an AI-assisted approach for annotating such image data and provide the four iterations of the annotation process and the final refined annotations together with the field images. The data is organized into folders following a train-validation-test split, where folders contain images and annotations of the splits. Python, 3.7.13 Ultralytics YOLOv5, Release v7.0 Germany Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy EXC-2070 - 390732324 (PhenoRob)