Egocentric videos : Social gathering analysis

In computer vision, video analysis and activity recognition are considered as a classic tasks. But activity recognition is relatively less well-defined as compared to some other neigh- boring problems such as object recognition. Action and activity recognition have been the subject of vast amount of research in computer vision field in the last decade. With the advent of wearable cameras, such as GoPro and Google Glass, there has been an increas- ing interest in egocentric vision that enables automated analysis of video captured from a first-person perspective. In my research work I am planning to handle and explore some challenges of video analysis and video summarization of different types of videos, such as videos captured by surveillance cameras and egocentric videos where video is captured by a wearable camera which can be mounted on a human, vehicle or animal. It can be further categorized as third person view point and first person view point respectively. I am open to investigate the problem of activity and action recognition in egocentric videos.


  1. Egocentric vision is a recent topic that inculcalte new problems and challenges produced by first-person perspective in video analysis.
  2. While first-person perspective generates additional challenges, also it posseses an enhanced field of view.
  3. The problems like analyzing social interactions, detection and tracking of objects, extracting salient moments or summarization from our daily activities are perceiving a new shape/angle.


Given egocentric videos, find group forming patterns to analyse type of social iteraction. In these videos the first-person includes himself/herself in the group and captures the videos


  • Egocentric vision
  • EgoTracking
  • Social interaction patterns