Over the past decade's person re-identification is being a hot research area and an active participant in the automated video surveillance. Besides monitoring of person of interest and human-machine interaction, person re-identification is also used in broadcast media and in forensic examination. Images and videos captured from different cameras gets suffers from low quality and resolution issues which generates difficulty in extraction of useful information. Deep learning plays a vital role in person re-identification in retrieval and determining similarity among the features from data. With the advancement in the deep learning techniques, favorable performances are obtained while handling challenges obtained from diverse viewpoints, dazedness, resolutions in image, settings of camera, blocking of an object of interest and irregular background across camera views. This paper presentes the current challenges that are overcome by the usage of the deep learning architectures.
Authors: Neha Mathur, Shruti Mathur, Divya Mathur, Pankaj Dadheech
DOI: https://doi.org/10.1109/icetce48199.2020.9091747
Publish Year: 2020
