One of the insect’s worst enemies — and a significant contributor to colony collapse — is a small mite called Varroa destructor that’s as devastating as it sounds. They latch onto bees and brood, sucking the life out of them.
A handful of solutions have emerged over the past few years, including hot hives that kill the mites and robobees that replace natural pollinators.
Now, a concerned team of beekeepers and software developers from Sweden have proposed another plan. They want to build an AI-powered app called BeeScanning that would analyze images of beehives to spot the mites, then alert keepers so they could rid their hives of the mites.
“I’ve been taking pictures of my bee colonies to learn how to foresee their development,” Björn Lagerman, a BeeScanning founder who’s been beekeeping for 45 years, told Digital Trends. “Last summer, I started taking pictures of frames with brood to learn the capacity of different queens. Examining the stills, I discovered mites, which I didn’t see on the moving bees when I first took the pictures.”
The mites appear as little red dots on the backs of bees. The more pictures Lagerman took, the more mites he found. He realized he could develop a tool that could scan the images much faster and more accurately than his own eyes.
Lagerman got together a small team and took to Kickstarter to seek funding. Initially, they plan to launch a tool to help keepers around the world collaborate by uploading images of their own hives.
The images these keepers upload will help the BeeScanning team train an AI to identify individual mites in a picture full of bees. By the end of 2017, Lagerman hopes to compile a database of 40,000 images from 10,000 hives, which would be used to train an algorithm to efficiently spot the mites. They want to raise money on Kickstarter to fund the app’s software development and distribution.
“If you don’t monitor varroa and treat accordingly, your bees will die,” Lagerman said. “That’s the motivation driving beekeepers to examine their colonies in all ways they can think of!”
The project reached its goal of 50,000 Swedish krona ($5,745) within the first 10 days of the campaign. That money will help fund the database. The team has now added a number of stretch goals to enable features such as offline mode and a community-focused web interface.
