Towards Recognition of Human Activities and in depth Analysis of Evaluation Datasets

Abstract

Human activity detection is one of the key areas of interest in all surveillance applications. This work has earlier been performed manually but recent technologies have successfully converted into automated systems to carry out this task. A variety of online dataset repositories are available for training the systems and testing them later. Some datasets are solely inclusive of single human actions, however, in some others, the interactions of humans with objects and other humans have also been accounted for. This paper focuses on single human actions alone and a survey of datasets with the dominance of single human actions. Keywords: