A formal approach to good practices in Pseudo-Labeling for Unsupervised Domain Adaptive Re-Identification

摘要

The use of pseudo-labels prevails to tackle Unsupervised Domain Adaptive (UDA) Re-Identification (re-ID) with the best performance. Indeed, this family of approaches has given rise to several UDA re-identification-specific frameworks, which are effective. In these works, research directions to improve Pseudo-Labeling UDA re-ID performance are varied and primarily based on intuition and experiments: refining pseudo-labels, reducing the impact of errors in pseudo-labels... It can be hard to deduce from them general good practices that we can implement in any Pseudo-Labeling method to improve its performance consistently. We propose a new theoretical view on Pseudo-Labeling UDA re-ID to address this fundamental question. The contributions are threefold: (i) A novel theoretical framework for Pseudo-Labeling UDA re-ID, formalized through a new general learning upper bound on the UDA re-ID performance. (ii) General good practices for Pseudo-Labeling are directly deduced from the interpretation of the proposed theoretical framework to improve the target re-ID performance. (iii) Extensive experiments on challenging person and vehicle cross-dataset re-ID tasks, showing consistent performance improvements for various state-of-the-art methods and various proposed implementations of good practices.