Ukraine has developed artificial intelligence to detect disguised Russian positions
The Ukrainian program Clarity, which analyzes drone images and automatically marks suspicious enemy objects on them, was presented at Brave1 Defense Tech Valley 2025.
"The military flies on photo flights like "Mara", "Valkyrie", Cetus. They take photos of the battlefield to a depth of 20 km. When the plane lands, it brings with it 1,500-2,000 high-quality images. It takes analysts up to 6 hours to process them manually. And we have created software that will allow you to do this in 20-30 minutes," a company representative told "Oborontsi".
The idea of Clarity is to give the analyst an initial analysis of the images. The analyst sees where to look first, and this helps him to form a report faster. And based on this report, the command can make a decision much faster about destroying the found object.
The neural network distinguishes between military and civilian equipment, camouflaged objects, and also notices traces that may indicate the presence of an enemy position nearby. If 10 objects are seen in one photo, and zero in most others, then the program itself suggests where to pay attention.
"We have our own dataset of 100,000 images, on which we trained a neural network to recognize objects from photos. We collected it based on open sources, including YouTube. And then we listened to the military, who told us what to improve.
The product is positioned as an auxiliary solution. That is, the final report is still generated by a person to avoid errors or omissions. But this tool saves valuable hours for decision-making.
Clarity is already used in the Unmanned Systems Forces, the Armed Forces of Ukraine, the National Guard, and the Border Guard Service. In the future, the developers plan to adapt the product for civilian tasks. For example, search missions, border monitoring, and tracking illegal logging.
Previously, "Defense" wrote that at Brave1 Defense Tech Valley 2025, the Logic7 company developed a virtual simulator for shooting down FPV drones with a shotgun.