Real-Time Human Action Recognition using Lightweight Deep Learning Networks

1Mr. K. Jayachandra, M.Tech,
Assistant Professor,Department of AIDS,Annamacharya Institute ofTechnology and Sciences,
Tirupati.-517520, A.P. .jayachandra502@gmail.com
4Hema Sandhya C
UG Scholar, Dept. of ArtificialIntelligence And Data Science,Annamacharya Institute of
Technology and Sciences,Tirupati, Indiahemasandhya5@gmail.com
2Jyoshna .Y
UG Scholar, Dept. of ArtificialIntelligence And Data Scinece,Annamacharya Institute of
Technology and Sciences,Tirupati, Indiajoshujoshu253@gmail.com

Mohammad Bilal C
UG Scholar, Dept. of ArtificialIntelligence And Data Science,Annamacharya Institute of
Technology and Sciences,Tirupati, Indiamohammadbilal200327@gmail.com
5Davoodh Shaik
UG Scholar, Dept. of ArtificialIntelligence And Data Science,Annamacharya Institute of
Technology and Sciences,Tirupati, Indiasdavoodhdavoodh@gmail.com

Abstract— Human Action Recognition (HAR) fromvideo streams is an important research area of computer vision. The applications of this research area includehealthcare monitoring, smart home, and tele-immersion. However, recognizing human actions is a challenging problem due to several reasons, including the appearance of humans and the occurrence of occlusions, lighting conditions, and complexity of backgrounds. The overall performance of the HAR system depends on an efficient feature extraction mechanism and the training of thesystem.