Person Re-Identification: A Retrospective Study, Recent Trends and Future Scope
Shashank Kumar Soni, Gautam Kumar · 2024
Person re-identification (Re-ID) searches several non-overlapping cameras for a person of interest. The development of deep neural networks combined with the increasing need for intelligent video surveillance has led to a huge surge in interest in the field of computer vision. This article highlights the state-of-the-art techniques used for person re-identification (reID), including the latest advancements and challenges. It outlines the different person re-identification datasets that are available in the literature, evaluates the effectiveness of the best techniques across a range of datasets, and clarifies the important restrictions on person re-identification, including the effects of changing environmental conditions and data limitations. Furthermore, the article points out the future research direction, focusing on creating more realistic datasets and designing robust algorithms that can effectively handle real-world conditions in order to drive the growth and development of re- ID technology.