Human action recognition based on MnasNet optimized by improved version of Football Team training algorithm

Human action recognition has applications in retrieval, human–machine interaction, and surveillance. In this paper, an innovative method for the recognition of human action from the images has been presented. Since human action is based on the placement of body parts and in each action, different parts of the body take on different meanings, it is very important to use the different parts of the body to identify human actions. In this article, a deep neural network, called MnasNet has been proposed in order to identify human action. The architecture of the MnasNet network in this study, has been modified by fine-tuning its parameters by an improved version of the Football Team Training (IFTT) algorithm. The suggested methodology integrates the advantages of MnasNet with the advanced FTTA, leading to enhanced accuracy and efficiency in recognition tasks. Experimental results on benchmark datasets validate the efficacy of the proposed MnasNet/FTTA model, accomplishing state-of-the-art results in the domain of human action recognition.