Using Segmentation Neural Networks for Accelerated Object Detection in an Image

Abstract

The purpose of this scientific work is to accelerate the process of detecting objects in the image, using segmentation neural networks for this. The research methodology is the use of open datasets to compare the speed and accuracy of the main architectures of neural networks for finding objects and the approach used in this article. As a result of the research, it became known that the developed algorithm speeds up finding objects in the image from twenty to two thousand times, without losing the accuracy of work.