Artificial intelligence elements for the task of determining the position of the vehicle in the image

The article is devoted to solving the problem of determining the boundaries of the vehicle on the image. This task is an intermediate step to solve other, more local tasks related to the identification of vehicles in the image or video stream. The article in detail considers existing methods and approaches to solving problems of computer vision, including modern architectures of neural networks. Tiny-YOLO-InceptionResNet was chosen as the primary neural network and was modified during the research process. The architecture of the resulting neural network is given in this paper. The training of the neural network was preceded by the preparation of a data set that allowed for a more rational use of computing resources during training. As a result of the research the model of finding the boundaries of the vehicle on the image was developed. The accuracy of this model is 88%.