Improving the Efficiency of Neural Network Methods for UAV Identification and Tracking in the Image Stream Based on the Synthesis of Training Data Sets

at present, research on the detection of unmanned aerial vehicles (UAVs) is of particular relevance. One of the most effective ways to detect UAVs is video detection using computer vision algorithms. The purpose of this work is to increase the efficiency of algorithms for recognizing and tracking unmanned aerial vehicles in video images. The result of the work was an increase in the quality of detection and an increase in the mAp metric from 0.683 to 0.7385.