On January 8, the Wiley Journal “Microscopy Research and Technique” published the article, “DLgram cloud service for deep-learning analysis of microscopy images.” In it, scientists from the NSU Laboratory of Deep Machine Learning in Physical Methods at the Institute of Intelligent Robotics presented a universal service they developed to recognize various types of numerous homogeneous objects. Their work was supported by the Russian Science Foundation grant No. 22-23-00951.
Andrey Matveev, Head of the Laboratory, provided more details about the service,
We presented the first version of the service in 2021. Then, thanks to the support of a grant from the Russian Science Foundation, we were able to improve it. This involved expanding the functionality and introduced new capabilities for analyzing recognized objects. Our Telegram service is now quite famous among microscopists and has more than 300 users in Russia and abroad. No special skills are required to work with it. A neural network frees a person from performing routine processes associated with counting and characterizing particles or objects in an image. This includes determining their quantity and concentration as well as maximum, minimum, and average sizes. The service operates in cloud mode, i.e. the Cascade Mask-RCNN neural network used does not work on the user’s computer, but on the graphic server of the Institute of Intelligent Robotics of NSU.
Anna Nartova, Senior Researcher at the Laboratory, added,
Any user can work with the DLgram service, it does not require any special knowledge or skills. It is enough to mark 10 objects in your image in a selected square and upload it to the Nanoparticles Telegram channel (https://t.me/nanoparticles_nsk). The image is picked up by the chatbot and goes to the Institute of Intelligent Robotics server. There, the neural network is trained and the objects in the image are recognized. This process only takes a few minutes. The user receives an image with recognized objects and can make adjustments as necessary. After this, the parameters of the objects are identified: quantity, size, area, and concentration.
The new service works with various photographic images, pictures from microscopes, cameras, and mobile phones, and the neural network recognizes not only particles of substances, but macro-objects.
Alexey Okunev, Director of the NSU Institute of Intelligent Robotics, explained that as part of the Federal project for the development of Artificial Intelligence Centers, “We are actively developing the creation of digital assistants in other fields. Employees and students at the Institute are working on creating automatic recognition systems for various objects in industry as well as for the urban environment”.
