Registration Number
0223U003611,
0121U109526
, R & D reports
Title
Real-time object localization, classification and tracking system using artificial intelligence algorithms
popup.stage_title
Head
Barkalov Olexandr O., Доктор технічних наук
Registration Date
31-05-2023
Organization
Vasyl' Stus Donetsk National University
popup.description2
The information system has been disassembled to dynamically integrate information about the context of the video and the positions of people based on the completeness and accuracy of information about the positions for better recognition of people. The novelty of the solution lies in the fact that the earlier methods of integration are more statically combining two data, or later enriching information about the posture after additional information about the external context, which leads to a strong shift in results peredbachen in the direction of the context. On the face of it, the integration of call signs is suggested, you can filter the contextual information and in this way you can win over the recognition of the sound on the video with the correct and the wrong context. In this way, it shows the prospects for integrating the technology of recognizing poses on video to improve the quality of work and other solutions. One of the problems that the solution is to recognize is the recognition of people in the video. Historically, the solution of this problem was placed on the meaning of the pixels themselves in the video, as they describe the context of the video, and only recently were proponated systems, like to analyze the context of the video with the optical flow of the video.
As part of the experiment, the idea of developing a neural network model for evaluating poses was proposed, as well as integrating the solution for recognizing people with a method of increasing the accuracy and quality of classifying people in videos.
As a result, the information system was split up, dynamically integrating information about the context of the video and people's positions in a new way based on the accuracy of information about the positions for better recognition of people. The novelty of the solution lies in the fact that the earlier methods of integration are more statically combining two data, or later enriching information about the posture after additional informat
Product Description
popup.authors
Baiev Artem V.
Barkalov Olexandr O.
Vetrov Oleg S.
Dyma Volodymyr S.
Komarov Vasyl F.
Lishchuk Andrii R.
Neskorodieva Anastasiia R.
Petranova Maryna Yu.
Strutovskii Maksym I.
Shtovba Serhii D.
popup.nrat_date
2023-05-31
