Ukraine and Russia train AI models from UAV data for combat

(CLO) As artificial intelligence (AI) is increasingly applied on the battlefield, both Ukraine and Russia are in possession of an invaluable resource: millions of hours of footage from unmanned aerial vehicles (UAVs). This data will be used to train AI models capable of making tactical decisions on the battlefield.

Both Ukraine and Russia have deployed AI in the conflict, particularly in identifying targets and analyzing imagery much faster than humans could.

Oleksandr Dmitriev, founder of OCHI, a Ukrainian non-profit digital system that collects and analyzes data from more than 15,000 frontline UAV crews, said his system has collected 2 million hours of battlefield video from drones since 2022. That's 228 years of data and will provide a valuable trove of data for AI to learn from.

"This is the 'food' for AI. If you want to teach AI, give it 2 million hours of video, it will become a supernatural tool," Dmitriev said.

Servicemen from a Ukrainian artillery unit are training to use drones. Photo: Dmytro Smolienko

Drone footage can be used to train AI models to recognize combat tactics, detect targets, and evaluate the effectiveness of weapons systems. “Experience can be turned into mathematics,” Dmitriev said. “An AI program can study the trajectories and angles at which weapons operate most effectively.”

The OCHI system was originally developed in 2022 with the aim of providing military commanders with an overview of the battlefield by displaying footage from UAVs of all nearby crews on a single screen.

After deployment, the OCHI team realized that these footages were not only useful for tactical management but also an important resource for recording the entire combat process, so they decided to archive these videos.

According to Dmitriev, the system collects an average of 5-6 terabytes of data from combat engagements every day, providing a huge amount of information to train the AI.

The quality of the images and the size of the dataset are important factors in helping AI identify targets accurately. Samuel Bendett, a senior fellow at the Center for a New American Security, said such large amounts of data would be invaluable in training AI systems to accurately identify targets and the tactical situations they face.

“Humans can visually recognize situations, but machines can’t. They have to be trained to recognize elements like roads, natural obstacles, or ambushes,” he explained.

Kateryna Bondar, a research fellow at the Wadhwani Center for AI at the Center for Strategic and International Studies, stressed that the quality of the images and the scale of the data are extremely important for AI to learn to recognize targets based on shape, color, and other features.

In addition to OCHI, Ukraine has another system called Avengers, developed by the Ministry of Defense, which also collects and analyzes video from UAVs and surveillance cameras. While the Ukrainian Ministry of Defense did not provide details about the system, it has previously revealed that Avengers is capable of detecting 12,000 Russian devices per week through its AI recognition engine.

In addition to using AI for data collection and target identification, Ukraine is also developing autonomous UAV technology. The drones can fly at targets without human control. Additionally, AI is being used in mine clearance missions, helping to reduce the risk to military forces and civilians.

Russia has also deployed AI on the battlefield, notably in the targeting of its Lancet attack UAVs, which have proven effective in attacking Ukrainian armored vehicles.

Hoai Phuong (according to Reuters, WP, TASS)

Source: https://www.congluan.vn/ukraine-va-nga-dao-tao-mo-hinh-ai-tu-du-lieu-uav-de-phuc-vu-chien-dau-post326670.html