Pixel and Feature Transfer Fusion for Unsupervised Cross-Dataset Person Re-Identification

Recently, unsupervised cross-dataset person reidentification (Re-ID) has attracted more and more attention, which aims to transfer knowledge of a labeled source domain to an unlabeled target domain. There are two common frameworks: one is pixel-alignment of transferring low-level knowledge and the other is feature-alignment of transferring high-level knowledge. In this paper, we propose a novel Recurrent Auto-Encoder (RAE) framework to unify these two kinds of methods and inherit their merits. Specifically, the proposed RAE includes three modules, i.e. a feature-transfer module, a pixel-transfer module and a fusion module. The feature-transfer module utilizes an encoder to map source and target images to a shared feature space. In the space, not only features are identity-discriminative, but also the gap between source and target features is reduced. The pixel-transfer module takes a decoder to reconstruct original images with its features. Here, we hope the images reconstructed from target features are in source-style. Thus, the low-level knowledge can be propagated to the target domain. After transferring both high- and low-level knowledge with the two proposed module above, we design another bilinear pooling layer to fuse both kinds of knowledge. Extensive experiments on Market-1501, DukeMTMC-ReID and MSMT17 datasets show that our method significantly outperforms either pixel-alignment or feature-alignment Re-ID methods, and achieves new state-of-the-art results.