A U-Net–Based Segmentation Approach for Honeybee Pollen Detection Using Deep Learning

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Pollen is the most important food source for maintaining both colony development and honey production in bee populations. Therefore, the presence of pollen in bee colonies is one of the biggest determinants of hive health. In this context, this study contributes to animal and agricultural sustainability by presenting a deep learning-based image analysis approach for the automatic detection of pollen carried by bees. The bee pollen recognition study was performed using the U-Net deep learning architecture applied to the Pollen Dataset. The U-Net model was selected due to its ability to preserve high-resolution spatial information between input images and corresponding segmentation maps, thereby improving detection accuracy. Experimental results show that the proposed model achieved a successful performance with an F1 score of 99.4%. In addition to technical performance, the study also highlights the potential applications of bee pollen detection in agriculture, particularly in supporting farmers, researchers, and policymakers with more informed decision-making processes. Overall, the findings indicate that advances in deep learning-based image analysis for pollen detection represent an important step towards promoting both environmental sustainability and agricultural resilience.