Assessing colony health is a deceptively detail-oriented and time-consuming task. Beekeepers and researchers alike must count the number of combs and assess the quality several times a year to gauge overall colony health. Considering the density and size of colonies, this process is labour-intensive and prone to error. Researchers Thiago da Silva Alves and Jean Metz from the Polytechnic Institute of Bragança invented DeepBee in 2018 to try to make assessing colony health more efficient. DeepBee is a tool that streamlines colony assessment by introducing machine learning and Deep Neural Networks to assess colony health through images. Images of a given colony’s honeycomb are segmented into smaller chunks and then classified based on the architecture (arrangement) of the combs. Alves and Metz identified 13 unique architectures and created algorithms for DeepBee to automatically identify and classify combs through image analysis. DeepBee is still in development but has been presented as a promising new innovation to make assessing colony health faster and possibly more accurate. With the number of bee colonies in rapid decline globally and considering the rate at which an entire hive can collapse, being able to better monitor colony health could provide more information on the timeline of events that lead to colony collapse while providing insight on what could be done to prevent collapse. It is worth considering how user-friendly DeepBee will be; would beekeepers need to provide samples to DeepBee to correctly analyze their specific colony correctly? And does the species of bee or climate influence the structure of the combs? DeepBee is still in development and will have to face these questions as the project buzzes closer to a marketable good.


Artificial Intelligence, Biodiversity, Ecological Monitoring, Industry/Natural Commodities, Visual Technologies