Temporal Activity Detection in Dense Video Captioning

Abstract

In today’s world recognizing activities has become an important task over the last few years due to continuous increase in the number of video devices and databases. This data needs to be analyzed and classified so that we know what the video is all about. This work provides the importance of temporal activity detection in dense captioning. Temporal activity detection is the process of detecting the activities in the long untrimmed video, Classify the activity and in the input video, locate each instance of activity Many algorithms already. Many algorithms already exist but the problem with the current algorithm is that they detect past and current events only and there by forgets the future events. Thus to solve this problem a new algorithm is proposed which works in both direction forward and backward such that not only past and current events will get detected but also future events also detected.