Review on recent Computer Vision Methods for Human Action Recognition

• Azhee Wria Muhamada ^[1] ; Aree A. Mohammed ^[2]
  1. [1] Computer Science Department, College of Basic Education, University of Sulaimani, Iraq
  2. [2] Faculty of Science, Computer Department, University of Sulaimani, Kurdistan Region, Iraq
• Localización: ADCAIJ: Advances in Distributed Computing and Artificial Intelligence Journal, ISSN-e 2255-2863, Vol. 10, Nº. 4, 2021, págs. 361-379
• Idioma: inglés
• Enlaces
  • Texto completo
• Resumen

  • Human action recognition has been an important goal of computer vision ever since its starting and has developed considerably within the last years. The recognition of human activities is sometimes thought of to be a straightforward method. Issues occur in advanced scenes involving high velocities. Activity prediction mistreatment of artificial intelligence (AI) by numerical analysis has attracted the eye of many academics. To modify the comparison of these ways, several datasets concerning tagged act created, having nice variation in content and methodology Human activities are a significant challenge in varied fields. There are several friendly applications during this space, as well as sensible homes, valuable artificial intelligence, human-computer interactions, and enhancements in protection in many areas like security, transport, education, and medication through the management of falling or aiding in medication consumption for older people. The advanced improvement and success of deep learning techniques in various pc vision applications encourage the utilization of those ways in the video process. The human presence is a fundamental challenge within the analysis of human behavior through activity. An individual in a more than video sequence may be represented by their motion, skeleton, and abstraction characteristics. This work aims to boost human presentation by gathering various options and, therefore, exploiting the new RNN structure for activities. Throughout this review, the paper focuses on recent advancements within the field of action recognition supported Machin learning. We have compared some of the triumphant human action recognition methodologies to accuracy and prediction along within the review paper