A Study of Deep Learning Models in Human Actions Recognitions (HAR) in Video Sequences

Exploring Deep Learning Models for Human Action Recognition in Video Sequences

Abstract

Identifying people in videos and figuring out how to fix the problem that comes with it. The mostpopular and significant state-of-the-art solutions are presented. The ConvNet topologies based on deeplearning to address the shortcomings of existing hand-coded methods, These ConvNets frameworks usea pre-trained deep model for features ex-tractions to recognize human behaviors in videos, making themsuitable for transfer learning. It is empirically shown that deep pre-trained model built on a big annotateddataset is ex-changeable to action recognition task using the smaller training dataset. a deeply linkedConvNet for human activity detection presented that utilize the RGB frames at the top layer with BidirectionalLong Short Term Memory (Bi-LSTM), and at the bottom layer, CNN model is trained using asingle Dynamic Motion Image (DMI).