Identity-Seeking Self-Supervised Representation Learning for Generalizable Person Re-identification

This paper aims to learn a domain-generalizable (DG) person re-identification (ReID) representation from large-scale videos $\mathbf{\text{without}}\mathbf{\text{any}}\mathbf{\text{annotation}}$. Prior DG ReID methods employ limited labeled data for training due to the high cost of annotation, which restricts further advances. To overcome the barriers of data and annotation, we propose to utilize large-scale unsupervised data for training. The key issue lies in how to mine identity information. To this end, we propose an Identity-seeking Self-supervised Representation learning (ISR) method. ISR constructs positive pairs from inter-frame images by modeling the instance association as a maximum-weight bipartite matching problem. A reliability-guided contrastive loss is further presented to suppress the adverse impact of noisy positive pairs, ensuring that reliable positive pairs dominate the learning process. The training cost of ISR scales approximately linearly with the data size, making it feasible to utilize large-scale data for training. The learned representation exhibits superior generalization ability. $\mathbf{\text{Without}}\mathbf{\text{human}}\mathbf{\text{annotation}}\mathbf{\text{and}}\mathbf{\text{fine-tuning,}}\mathbf{\text{ISR}}\mathbf{\text{achieves}}\mathbf{\text{87.0\%}}\mathbf{\text{Rank-1}}\mathbf{\text{on}}\mathbf{\text{Market-1501}}\mathbf{\text{and}}\mathbf{\text{56.4\%}}\mathbf{\text{Rank-1}}\mathbf{\text{on}}\mathbf{\text{MSMT17}}$, outperforming the best supervised domain-generalizable method by 5.0% and 19.5%, respectively. In the pre-training$\to$fine-tuning scenario, ISR achieves state-of-the-art performance, with 88.4% Rank-1 on MSMT17. The code is at \url{https://github.com/dcp15/ISR_ICCV2023_Oral}.