MOSCOW, December 24. /TASS/. Russia’s student startup DronOko (DroneEye) has developed a computer vision technology for unmanned aircraft systems (UAS), which enables real-time data processing and object identification, the press service of the National Technology Initiative (NTI) Platform told TASS.

"A student startup Dronoko has developed a computer vision technology for UAS. The solution will enable real-time data processing and object identification onboard. The project automates search and control technologies, saving time for area monitoring. The developers' calculations suggest that the search automation already gives a 20-30% time-saving effect," the statement says.

The developed technology is already capable of saving approximately 1.5 to 2.5 hours when viewing, for example, an 8-hour video. It is planned in the future to speed up the viewing process 1.3 to 1.7 times. The press service added that the technology can be used at construction sites, in restricted-access industrial areas, and in the search for people and lost animals.

"It’s important to note that the solution has already undergone practical testing, including simulation-based testing, for example, in forested areas. The system’s high accuracy is based on giant dataset model training on databank with hundreds of thousands of aerial photographs. The current capability of the system makes it possible to detect people in the visible spectrum, as well as demonstrates its reliability and readiness for use at industrial facilities or in agriculture," said Yaroslav Seliverstov, a leading AI expert at University 2035.

The startup’s founder, Vladimir Ryasnoy, a student at the Institute of End-to-End Technologies at Don State Technical University (DSTU), said that the startup planned to train the neural network to work with infrared thermal imaging, which will allow the innovation to be used to search for people at night, in smoky areas, and under trees. In addition, the NTI press service reported, the team's next step will be to optimize the system for use on a small computing module mounted on a drone, allowing the device to operate in conditions without communication or Internet access. The team also plans to upgrade the system so as to enable it to identify and "hold" a person in motion between frames, as well as add the function for the fixation of the detection location on a map with approximate coordinates for providing prompt assistance.