Russian military forces have begun using a new version of the Geran-2 strike drones with a camera, radio control system, and machine vision system based on foreign components.
Thanks to the solutions implemented in the design of our mass-produced UAV, it can strike targets at a depth of up to 150 km with target recognition in flight, and can also be controlled by an operator using the FPV drone principle.
The machine vision module is based on the NVIDIA Jetson Orin, a modern single-board computer with a six-core ARM Cortex A78E processor running at frequencies up to 1.7 GHz, and 1.5 MB of L2 cache and 4 MB of L3, a graphics subsystem of the Ampere architecture (512 or 1024 CUDA cores and 32 tensor cores).
Power consumption scales from 7 to 25 W. The Jetson software ecosystem includes optimized libraries for computer vision and deep learning tasks. The platform supports all popular machine learning frameworks — TensorFlow, PyTorch, ONNX. The Geran-2 in the modernized version carries up to 90 kg of explosives.
Earlier, www1.ru reported that "the 'barbecue' for the 'Fagot' is no obstacle: the Soviet ATGM has become a destroyer of armored vehicles".
Read more materials:
"Patriot Park" confused An-12 and An-26 aircraft at the exhibition
The first flight tests of a new generation airship were conducted in Russia
Loading of RBK-500 cluster bombs with UMPC on the pylons of the Su-34 was shown in the photo
