UAV Protection:
two products
for your task

Fig. 1. Multi-target mode: five UAVs are simultaneously detected and tracked by the neural network within the area of responsibility against a complex cloudy background. Each target is marked with its own red detection box. On the left are the previews of cameras No. 1–4 and the "Detection History" log.

Fig. 2. The main screen of the "Anti-Drone" Software. A UAV is captured in a detection box; the turret's aiming crosshairs (yellow — target, red — current) are converging on the target. The left panel shows the previews of cameras No. 1–4 and the "Detection History" log. At the top is the system telemetry (camera FPS, neural network FPS, number of dropped frames).

Fig. 3. UAV recognition against a complex background: a residential house, greenhouses, a tree line, and a grey cloudy sky. The neural network reliably marks the target with a red box; the telemetry shows 9.72 fps from the camera and 2.51 fps of neural network processing.

Fig. 4. Detection of a low-visibility target at long range. A UAV only a few pixels in size is reliably detected by the neural network. At the bottom are the service parameters of the PTZ camera (Pan/Tilt/Zoom): P028/T008 Z001.

Fig. 5. Real-time target tracking. The aiming crosshairs are converged on the UAV, and the detection box stays on the maneuvering target. Stationary objects (a power line pylon, trees) are correctly filtered out as false positives.

Fig. 6. Field deployment of the standard workstation module of the "Anti-Drone" Software. It is delivered as a deb package and deployed on a standard laptop with an NVIDIA GPU directly at the protected facility — without a server rack and without major installation work.