Unity Style Transfer for Person Re-Identification

TL;DR

This paper introduces UnityStyle, a style transfer method using UnityGAN to generate style-unity images, reducing style disparities in person re-identification across different cameras, leading to improved matching accuracy.

Contribution

The paper proposes UnityStyle, a novel style transfer approach with UnityGAN for creating style-unity images, enhancing style robustness in person re-identification models.

Findings

Outperforms existing methods on benchmark datasets

Produces more style-robust features for re-identification

Reduces image artifacts caused by camera differences

Abstract

Style variation has been a major challenge for person re-identification, which aims to match the same pedestrians across different cameras. Existing works attempted to address this problem with camera-invariant descriptor subspace learning. However, there will be more image artifacts when the difference between the images taken by different cameras is larger. To solve this problem, we propose a UnityStyle adaption method, which can smooth the style disparities within the same camera and across different cameras. Specifically, we firstly create UnityGAN to learn the style changes between cameras, producing shape-stable style-unity images for each camera, which is called UnityStyle images. Meanwhile, we use UnityStyle images to eliminate style differences between different images, which makes a better match between query and gallery. Then, we apply the proposed method to Re-ID models, expecting to obtain more style-robust depth features for querying. We conduct extensive experiments on widely used benchmark datasets to evaluate the performance of the proposed framework, the results of which confirm the superiority of the proposed model.