A person re-identification system based on Swin-transformer network

Person re-identification is an important research direction in the field of computer vision, and has extensive application values. This paper proposes a person re-identification system based on Swin-Transformer, which integrates a semiautomatic data preprocessing method based on YOLOv8, a pedestrian feature extraction network based on SwinTransformer, and the efficient indexing mechanism of Faiss, so as to realize a comprehensive optimization solution for feature extraction and retrieval. The experimental results show that the system performs excellently in terms of similarity distribution, retrieval performance and matching confidence. The similarity distributions between positive and negative samples are obviously separated, and the Top-20 retrieval accuracy rate is as high as 87%. The system demonstrates excellent recognition accuracy and robustness in complex scenarios, providing strong support for the practical application of person re-identification.