Learning to Align via Wasserstein for Person Re-Identification

Existing successful person re-identification (Re-ID) models often employ the part-level representation to extract the fine-grained information, but commonly use the loss that is particularly designed for global features, ignoring the relationship between semantic parts. In this paper, we present a novel triplet loss that emphasizes the salient parts and also takes the consideration of alignment. This loss is based on the crossing-bing matching metric that also known as Wasserstein Distance. It measures how much effort it would take to move the embeddings of local features to align two distributions, such that it is able to find an optimal transport matrix to re-weight the distance of different local parts. The distributions in support of local parts is produced via a new attention mechanism, which is calculated by the inner product between high-level global feature and local features, representing the importance of different semantic parts w.r.t. identification. We show that the obtained optimal transport matrix can not only distinguish the relevant and misleading parts, and hence assign different weights to them, but also rectify the original distance according to the learned distributions, resulting in an elegant solution for the mis-alignment issue. Besides, the proposed method is easily implemented in most Re-ID learning system with end-to-end training style, and can obviously improve their performance. Extensive experiments and comparisons with recent Re-ID methods manifest the competitive performance of our method.

I. Introduction

Person re-identification [1]–[4] has received considerable academic attention, owing to its tremendous potential in security applications such as people tracking [5] and behavior analysis [6]. The essential objective of person Re-ID is to identify target persons from a large number of person images. These images are taken under various viewpoints, different illuminations, and human poses, which leads to a large intra-personal variation severely hindering the identification. The recent success of person Re-ID models stems from the deep convolutional neural networks [7]–[9]. Due to its convincing discriminative ability, the deep leaning methods provide a powerful person image representation that can partly alleviate these issues and hence show the superiority.