Russian scientists create neural network that recognizes images taken by microscope

At the Institute of Intelligent Robotics at Novosibirsk State University (NSU), a set of free online services based on the artificial intelligence iOk has been developed and launched for the automatic analysis and description of images taken with microscopes. The platform aims to relieve scientists from the routine tasks associated with manual image processing, as reported by the press service of the Ministry of Education and Science of the Russian Federation to "Gazeta.Ru."

"Scientists often face the task of characterizing images obtained from a microscope, such as determining the average size of objects or their quantity. Previously, scientists had to perform these manipulations manually, requiring a lot of effort and time," said Andrey Matveev, head of the Deep Machine Learning Laboratory at the Institute of Intelligent Robotics at NSU.

The iOk platform consists of three online services based on artificial intelligence (AI) for image analysis: No Code ML, DLgram, and ParticlesNN. The first two are implemented in the Telegram messenger as bots, and the third is a web service. The core of all programs is the Cascade Mask-RCNN neural network trained on 5,000 objects.

All three programs can work with various types of images, including snapshots from electron microscopes and digital cameras (including smartphones), as well as videos. They can recognize various objects, such as nanoparticles, microorganisms, and cells.

According to Matveev, NSU's services differ from analogs in that they do not require high-resolution images. Additionally, the Russian development can recognize and ignore noise and glare on images, which other AIs might have identified as separate objects, affecting the accuracy of reports.

"For the convenience of users, we have combined them on one iOk platform. The total number of its users is already more than 500 specialists," added Andrey Matveev.