Generalizable Person Re-identification Without Demographics

YiFan Zhang, Feng Li, Zhang Zhang, Liang Wang, Dacheng Tao, Tieniu Tan

Keywords: Generalizable Person Re-Identification, Distributionally robust optimization

Abstract: Generalizable Person Re-Identification (DG ReID) aims to learn ready-to-use cross-domain representations for direct cross-data evaluation. It typically fully exploit demographics information, e.g. the domain information and camera IDs to learn features that are domain-invariant. However, the protected demographic features are not often accessible due to privacy and regulation issues. Under this more realistic setting, distributionally robust optimization (DRO) provides a promising way for learning robust models that are able to perform well on a collection of possible data distributions (the ``uncertainty set”) without demographics. However, the convex condition of KL DRO may not hold for overparameterized neural networks, such that applying KL DRO often fails to generalize under distribution shifts in real scenarios. Instead, by applying the change-of-measure technique and the analytical solution of KL DRO, we propose a simple yet efficient approach, Unit DRO. Unit DRO minimizes the loss over a reweighted dataset where important samples (i.e. samples on which models perform poorly) will be upweighted and others will be downweighted. Empirical results show that Unit DRO achieves superior performance on large-scale DG ReID and cross-domain ReID benchmarks compared to standard baselines.

One-sentence Summary: Introduce a novel setting in the Person ReID, that is learning invariant features without demographics. We reformulate the DRO problem as an importance sampling problem, termed Unit DRO, achieving superior performance compared to standard baselines.