In Ukraine, they tested a “smart” kamikaze UAV, which itself determines its targets

In Ukraine, they tested a “smart” kamikaze UAV, which itself determines its targets

The Ukrainian drone manufacturing company Wild Hornets is testing “smart” drones with a neural network and a target tracking system powered by Ukrainian electronics. This is stated on the official community page on the X social network.

As the manufacturers write, the neural network is located on a Ukrainian-made microcomputer measuring 30x30, built into the UAV. At the same time, the tested drone itself consists entirely of Ukrainian-made parts: flight controller, engine governor, video transmitter, camera, microcomputer, radio receiver (ELRS).

“Using a smart system, he himself identifies targets on the battlefield (vehicles, infantry). The operator can only select one of the identified targets and select it by pressing a button on the control panel,” note the authors of the publication.

In the video released by the manufacturers, you can see how the drone with machine vision recognizes people and vehicles, both static and in motion. The system highlights potential targets, highlighting them on the screen with red or green frames. In addition, the drone demonstrates the ability to independently evade objects flying in its direction, likely making it more difficult for invaders to shoot it down during flight.

Wild Hornets test smart drones with neural network on Russian electronics

In the video, the drone has a neural network and target tracker. Using the tool for smart system, itself determines the targets on the battlefield (vehicles, infantry). The operator only needs… pic.twitter.com/qXBQa0SfIn

— Wild hornets (@wilendhornets) April 21, 2024

The Defence Express portal writes that the main advantage of drones with machine vision is the ability to process video images, recognize the outlines of objects and attack them without human intervention. This technology eliminates the dependence of the effectiveness of an attack on the skill of the operator, and also eliminates the dependence on electronic warfare, since the drone can overcome the last conditional 100 meters on its own.

Experts also emphasized that the technology is not know-how, but it is necessary for it to function even on a cheap compact microcomputer. In addition, there are high requirements for the software itself, and the neural network itself needs to be trained to recognize objects under different conditions.