Improving the accurate answer about human actions in the visual question and answer system

Abstract

Activity detection is a necessary tool that enables efficient analysis of human behavior as well as efficient interactions between humans and vision systems. At the same time, recognition of human action is a very challenging task due to significant changes in human actions, in terms of personal styles, human appearance, camera view, changing background and other environmental changes, but so far an effective and efficient model that can recognize with few steps calculations and reveal the understanding of the image, is not presented. This field of research has attracted the attention of many researchers due to its application in various fields such as medicine, human-computer interaction, business, etc. For this purpose, with the aim of improving the accuracy of answers in the image question and answer system and increasing the accuracy of human activity detection, we have developed a solution of using deep learning mechanism to recognize the information of images and the image question and answer system to predict the answers to the questions in We suggest pictures. Our proposed method scales, deploys and monitors the ring detection framework. Finally, the answer to the question is calculated according to the weights calculated in the training stage, the features of the input image and the basis of correctness, and finally, according to the proposed clustering, it leads to an improvement in the accurate answer. The proposed method is superior in accuracy and speed of algorithm execution compared to other methods.