Cross-Modality Earth Mover’s Distance for Visible Thermal Person Re-identification

Visible thermal person re-identification (VT-ReID) suffers from inter-modality discrepancy and intra-identity variations. Distribution alignment is a popular solution for VT-ReID, however, it usually restricted to the influence of In this paper, we propose Cross-Modality Earth Mover's Distance (CM-EMD) that can alleviate impact variations during modality alignment. CM-EMD selects an optimal transport strategy assigns high weights pairs have smaller variation. manner, model will focus on reducing while paying less attention variations, leading more effective Moreover, introduce two techniques improve advantage CM-EMD. First, Discrimination Learning (CM-DL) designed overcome discrimination degradation problem caused by By ratio between inter-identity variances, CM-DL leads learn discriminative representations. Second, construct Multi-Granularity Structure (MGS), enabling us align modalities both coarse- fine-grained levels with proposed Extensive experiments show benefits its auxiliary (CM-DL MGS). Our method achieves state-of-the-art performance VT-ReID benchmarks.