STHARNet: spatio-temporal human action recognition network in content based video retrieval

Most of the needed information is easily accessible from our fingertips using the internet. The search procedure via the internet is a tough task behind the scenes. Content-Based Video Retrieval (CBVR) is one such search procedure on the internet. Human action recognition is one of them in CBVR. Even though there is research in these areas, the challenges are also partially solved. This paper also addresses the issues in human action recognition by designing an architecture named the Spatio-Temporal Human Action Recognition Network (STHARNet). The proposed STHARNet system is integrated into the CBVR system. The performance of the proposed architecture is evaluated by testing on three publicly available datasets: UCF Sports, KTH, and HMDB51. The results of the proposed architecture are encouraging and better than other recent methods.