Towards accurate person re-identification by deep learning

Artificial intelligence surveillance has become increasingly popular due to its security applications. Within this field, person re-identification (re-ID) is a. crucial topic, which aims at matching images of a person in one camera with the images of this person from other cameras. Considering the intensive appearance change of images of the same person, such as lighting, pose and viewpoint, person re-ID is a very challenging problem. In this thesis, we advocate addressing person re-ID by deep learning based methods, which have shown a much better representation ability and much stronger robustness to input data variation and corruption, compared to the traditional approaches. This thesis covers a series of problems involving re-ID including imagebased person re-ID and video-based person re-ID. We start by providing an overview of person re-ID. Following this, we show how to address different reID tasks accurately and efficiently, and our efforts have led to top-performing algorithms on all tasks. The thesis will conclude by describing several promising directions for future research.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2020

Author dc:creator

• Fu, Yang

Contributors dc:contributor

• Huang, Thomas S

• Person re-ID
• Deep Learning

Statement dc:rights

• Copyright 2020 Yang Fu

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/108109

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/108109