Review of Current Methods for Re-Identification in Computer Vision.

Abstract

The problem of reidentification of a person in multiple cameras is a hot topic in computer vision research. The issue is with the consistent identification of a person in multiple cameras from different viewpoints and environmental conditions.  Many computer vision researchers have been looking into methods that can improve the reidentification of people for many real-world purposes.  There are new methods each year that expand and explore new concepts and improve the accuracy of reidentification.  This paper will look at current developments and the past tends to find what has been done and what is being done to solve this problem.  This paper will start off by introducing the topic as well as covering the basic concepts of the reidentification problem.  Next, it will cover common datasets that are used in today's research.  Then it will look at evaluation techniques.  Then this paper will start to describe simple techniques that are used followed by the current deep learning techniques.  This paper will cover how these techniques are used, what are some of their weaknesses and their strengths.  It will conclude with an overview of some of the best models and show which models have the most promise and which models should be avoided.