Development of a self-learning model of collection and processing of information on a graphic processor for recognizing multi-parameter objects

Ignatiev Denis

Automatic object recognition is a demanded task for many industries, from military intelligence to manufacturing. The ability to visually monitor the environment without an observer has the potential to increase productivity. The purpose of this work is to develop a self-learning model for collecting and processing information on a graphics processor for recognizing multiparameter objects. The results showed that the GPU-assisted deep network implementation provides a fast solution for general object recognition. The lack of inherent properties of the system makes it impractical for most industrial applications. However, it should also be noted that none of the initial properties of the system make it particularly convenient for finding patterns in images, unlike other input spaces. In particular, nothing in the system uses the input space structure that comes with the use of images. Potential improvements can be exploited by integrating other techniques used for image recognition into a constrained Boltzmann machine or deep network. Thus, the proposed system does an excellent job with the assigned tasks and can be useful in useful areas.

Keywords:automatic object recognition, neural network, self-learning model, graphics processor, multi-parameter objects