Automatic detection of vehicles in video scenes obtained from drones

Abstract

In the qualification work, the development of the service of automatic detection of vehicles on video scenes was investigated. The training process of the YOLO model on the UAVDT dataset is shown, as well as the development process of the Nvidia Deepstream SDK service. The use of the Nvidia Deepstream SDK is established and justified due to its numerous advantages over other frameworks, as well as Yolov8's advantages over other neural networks.