infocus #66 June 2022 Fake honey, Machine Learning and Microscopy

At a conference I bumped into an eccentric fellow student who was toying with the idea of developing a new method for honey authentication.

His background was in machine learning, but he had little microscopy or bee research experience and so I offered to help. Honey is one of the world’s most faked products, and one can theoretically identify the origin of the honey from the morphology and size of the pollen in the honey.

Since current methods are either ineffective or prohibitively expensive, and faked honey harms both beekeepers and bees, an authentication tool that was affordable but effective would have tangible benefits. We also came to the realisation that it could have practical benefits in bee research and environmental monitoring. Indeed, when I reached out to a former supervisor who works on both honeybees and bumblebees, it transpired that all pollen-based research in bee research was being done manually.