Problem

The effective work of UAV-based recon and strike units on the frontline is one of the key advantages of the Armed Forces of Ukraine over the Russian occupation forces. The work of these units determines how quickly artillery, tanks, heavy equipment, EW, air defense equipment or even enemy headquarters will be detected and eliminated.

As part of such tasks, UAV recon operators have to review and process a large amount of information with small details with a limited time. Nevertheless, it’s often the key to the success of the entire operation.

Solution

To automate the work of UAV operators, DRONEPRO team created its own multilayer neural network. We «taught» it to recognize different types of enemy targets and create attack scenarios on various objects.

This system is based on deep learning, which allows it to improve results based on manual analysis of large numbers with annotations (examples).

Our team has processed tens of thousands of images in close cooperation with recon operators and we continue to do it every day. Thanks to this, the neural network can take into account many factors, such as size, shape, colors, etc., which allows it to obtain the most detailed information about the enemy target and achieve high accuracy and flexibility in recognizing different types of these targets.

Result

Currently, our AI target recognition system is used to automatically detect targets in UAV camera images or videos. In fact, it’s a detailed algorithm for processing images and analyzing data received from the camera. The result of his work is a list of detected targets with relevant characteristics such as size, shape and coordinates - in real time.