| Autor: |
Beknazarova, Saida, Abdullayev, Safibullo, Abdullayeva, Ozoda, Abdullayev, Zafarbek |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 2024, Vol. 3244 Issue 1, p1-7, 7p |
| Abstrakt: |
The aim of this article is to develop and test algorithms to recognize people and the tools they deal with specific times based on video. As part of this study, an algorithm consisting of several stages was proposed and implemented as a basic solution: recognition of people in the video frameand determination of the coordinates of the rectangularedge points where the person is located; determination of the coordinates of the key points of detected people in the video frame; recognition of tools in the video frame and their edge points Determine the tool in which the person is working at acertaintime (the time is calculated by the frame number from the video). To implement the algorithm, additional training of the existing computer vision model for then ext computer vision task was carried out a study that wastested. Object detection and definition of important points of people, especially people (poseestimation),objectoverlay. The following indicators were used asmetrics formulti-cluster tasks, which determine the toolsin the hands of a person in each frame (objectoverlay): accuracy, sensitivity. The algorithm was launched ontheweb service and tested by specialists. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
