Person re-identification for long term outfit problem

• Main Author: Wei, Zilai
• Other Authors: Alex Chichung Kot
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering; Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/166357
• Institution: Nanyang Technological University

Person re-identification (Person Re-ID) is a computer vision task that aims to detect and recognize the same person across different surveillance videos. Specifically, its objective is to match people with the same identity in a surveillance scene. This dissertation proposes a long-term clothing-changing person re-identification (Re-ID) method, which addresses the challenge of identifying individuals who usually change their clothing over long periods of time. Our approach includes applying three trained Self-Correction for Human Parsing (SCHP) extractors to segment the body parts of the person, and use pixel sampling to present for each label, then apply the Semantic-guided Pixel Sampling (SPS) model to train. The proposed method is evaluated on three publicly available datasets. Besides, the Baseline model based on ResNet50 is applied as the contrast method. The experimental results demonstrate that our approach outperforms the baseline model and some other state-of-the-art methods in terms of accuracy and robustness in identifying long-term clothing-changing individuals.