ANALYSIS OF ALGORITHMS FOR READING OBJECTS OF INTERFERENCE BY TELEPRESENCE ROBOT

Abstract

In this paper, we propose the development of a telepresence robot for object recognition. To do this, the authors get acquainted with different reading methods, their image processing speed and accuracy of reading other things and creatures in the environment they provide, then compare and choose the most optimal algorithm for different parameters. The goal is to develop software that allows telepresence robots to read objects of possible interference. The article describes and briefly describes the algorithms for touching the primary SSD model as Fast R-CNN and YOLO. A general description of the SSD model is given. It has also been described in more detail as an SSD model. The process of image processing and the stage of learning the functional model is provided. It was also explained why a solid-state drive is the best model in terms of accuracy and speed, even if the input size of this model is much smaller than the input size of its direct competitor - the YOLO model. In addition, there was a difference in a model building between the two object recognition models. It was described in detail the stage of learning the functional model, what formulas are used in the calculations and what they affect.