Survey on Reliable Deep Learning-Based Person Re-Identification Models: Are We There Yet?

Abstract

Intelligent video-surveillance (IVS) is currently an active research field in computer vision and machine learning and provides useful tools for surveillance operators and forensic video investigators. Person re-identification (PReID) is one of the most critical problems in IVS, and it consists of recognizing whether or not an individual has already been observed over a camera in a network. Solutions to PReID have myriad applications including retrieval of video-sequences showing an individual of interest or even pedestrian tracking over multiple camera views. Different techniques have been proposed to increase the performance of PReID in the literature, and more recently researchers utilized deep neural networks (DNNs) given their compelling performance on similar vision problems and fast execution at test time. Given the importance and wide range of applications of re-identification solutions, our objective herein is to discuss the work carried out in the area and come up with a survey of state-of-the-art DNN models being used for this task. We present descriptions of each model along with their evaluation on a set of benchmark datasets. Finally, we show a detailed comparison among these models, which are followed by some discussions on their limitations that can work as guidelines for future research.

Person Re-Identification with RGB–D and RGB–IR Sensors: A Comprehensive Survey

“…In this survey, we include only IR–IR and depth–depth single-modal person Re-ID because previous surveys did not include these types of single-modal Re-ID methods [ 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 ]. They mainly focused on RGB–RGB single-modal person Re-ID methods.…”

“…Therefore, we aim to provide a quick understanding and overview of the Re-ID researchers who are interested in working with depth and infrared sensors. There are several remarkable person Re-ID surveys that exist in the literature [ 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 ]. However, these surveys mainly focus on and give insightful direction on RGB camera-based person Re-ID approaches.…”

“…All of these modality-aware, Re-ID works consist of conventional methods [ 14 , 34 , 35 , 36 ] or deep-learning methods [ 37 , 38 , 39 ]. Because all the previous surveys [ 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 ] mainly focus on RGB–RGB single-modality matching, we do not include it in our survey.…”

Deep learning-based person re-identification methods: A survey and outlook of recent works

“…Prior to this survey, some researchers [4,5,42,44,46,46,47,81,111,116,118,121,127,131,137,150] have also reviewed person Re-ID field. In Table 1, we summarize the major contributions of these reviews.…”

“…Some of these surveys [5,150] summarized image-based and video-based person Re-ID methods. Other surveys [46,81,111,121,131,137] summarized deep learning-based person Re-ID method in different dimensions, which has developed rapidly after 2014 and become the main research means. Recently, Wang et al [116] have outlined methods of cross-domain person Re-I and compared the performance of these methods on public dataset.…”