Abstract:
Action recognition is one of the most important fields in computer vision. Hence, there is an open question of the high accuracy of complex background of human activities. A deep learning approach has recently been used to increase recognition validity with different application areas such as video surveillance, entertainment, autonomous driving vehicles, and human–machine interactions, etc. The aim of this research is to recognize human religious actions that differ in different activities. In our study, we have created our dataset from religious praying videos collected from YouTube, which has been classified into four different classes in terms of religion. We have applied a deep convolutional neural network using the Resnet-50 model for identifying human activity recognition (HAR) and we have got 98.79% accuracy. This research will help to cover more human action recognition tasks of daily activities.
