Artificial Intelligence is a technology that promises to solve many problems, but it doesn’t seem likely that it would help preserve our honeybee population. As unexpected as it seems, though, AI can be used to protect bees from the Varroa mite and will help beekeepers keep their hives healthy and safe for the declining honey bee population.
The Varroa mite is a tiny bloodsucking parasite that feeds on bees, both adult and young. It will weaken the colony, resulting in deformed and weakened generations of bees. While the parasite is not a direct cause of death for the bees, their presence will eventually lead to a complete collapse of the colony.
The mites are so tiny that they will escape detection by the naked eye. Only a millimeter in length, the colony can be affected for a very long time before the devastating results of the infestation are noticed.
Historically, beekeepers have had to carefully pick through the leavings of the hive, collected in a flat pan under the hive itself, to find the bodies of the tiny terrors if they wanted to stop the infestation. Beekeepers try to meticulously comb through hive dirt and waste to find Varroa mites, and it is easy to miss the tiny creatures and not realize there is a problem.
AI is far superior to the naked eye in picking up undetectable data in a noisy background, like finding dead Varroa mites in piles of hive dirt, and in fact, students in Switzerland have invented ApiZoom to do just that.
École Polytechnique Fédérale de Lausanne students have been working on ApiZoom, programming it to detect and identify the tiny mites in pictures. Beekeepers feed the data to the application by uploading photos to the EPFL site via smart device.
The students started testing the application in 2017 and have a 90% success rate. The ApiZoom model has been trained to thousands of images, and the success rate is similar to humans performing the same task.
On the heels of this success, according to project member Alain Bugnon, the students are planning on widely distributing the app as soon as possible. The application will be unrolled in two phases, first a web application to be followed by a smartphone application.
By combining the findings of these two approaches, it will be possible to evaluate the degree of infestation in individual hives as well as hives across a region. In addition, the app will be able to identify any possible mutations of the mites and any beekeeper practices that might unintentionally encourage an infestation.
An appropriate infestation response can be coordinated using the data collected from ApiZoom. Alain Bugnon is planning on releasing ApiZoom as its own company, which will speed the production and release of the software to beekeepers.