Person Re-Identification (ReID) task seeks to enhance the tracking of
multiple individuals by surveillance cameras. It provides additional support
for multimodal tasks, including text-based person retrieval and human matching.
One of the primary challenges in ReID is clothes-changing, which means the same
person wears different clothes. While previous methods have achieved
competitive results in maintaining clothing data consistency and handling
clothing change data, they still tend to rely excessively on clothing
information, thus limiting performance due to the dynamic nature of human
appearances. To mitigate this challenge, we propose the Pose Guidance by
Supervision (PGS) framework, an effective framework for learning pose guidance
within the ReID task. This approach leverages pose knowledge and human part
information from the pre-trained features to guide the network focus on
clothes-irrelevant information, thus alleviating the clothes' influence on the
deep learning model. Extensive experiments on five benchmark datasets
demonstrate that our framework achieves competitive results compared with other
state-of-the-art methods, which holds promise for developing robust models in
the ReID task. Our code is available at this https URL

通过 Pose Guidance by Supervision（PGS）框架，利用预训练特征中的姿势知识和人体部位信息引导深度学习模型关注与服装无关的信息，从而缓解服装对模型的影响，进而在 Person Re-Identification 任务中获得了具有竞争力的结果。

姿势引导监督：服装更换人员再识别的框架

Pose Guidance by Supervision: A Framework for Clothes-Changing Person  Re-Identification

Clothes-invariant feature extraction is critical to the clothes-changing
person re-identification (CC-ReID). It can provide discriminative identity
features and eliminate the negative effects caused by the confounder--clothing
changes. But we argue that there exists a strong spurious correlation between
clothes and human identity, that restricts the common likelihood-based ReID
method P(Y|X) to extract clothes-irrelevant features. In this paper, we propose
a new Causal Clothes-Invariant Learning (CCIL) method to achieve
clothes-invariant feature learning by modeling causal intervention P(Y|do(X)).
This new causality-based model is inherently invariant to the confounder in the
causal view, which can achieve the clothes-invariant features and avoid the
barrier faced by the likelihood-based methods. Extensive experiments on three
CC-ReID benchmarks, including PRCC, LTCC, and VC-Clothes, demonstrate the
effectiveness of our approach, which achieves a new state of the art.

本文提出了一种基于因果干预建模的 Causal Clothes-Invariant Learning (CCIL) 方法，可以实现穿衣不变的特征学习。在三个 CC-ReID 基准测试上进行了广泛的实验，证明了该方法的有效性，取得了新的最先进技术水平。