Clustering-based approach has proved effective in dealing with unsupervised domain adaptive person re-identification (ReID) tasks. However, existing works along this approach still suffer from noisy pseudo labels and the unreliable generalization ability during the whole training process. To solve these problems, this paper proposes a new approach to learn the feature representation with better generalization ability through limiting noisy pseudo labels. At first, we propose a Sample Dropout (SD) method to prevent the training of the model from falling into the vicious circle caused by samples that are frequently assigned with noisy pseudo labels. In addition, we put forward a brand-new method referred as to Feature Diversity Learning (FDL) under the classic mutual-teaching architecture, which can significantly improve the generalization ability of the feature representation on the target domain. Experimental results show that our proposed FDL-SD achieves the state-of-the-art performance on multiple benchmark datasets.
翻译:事实证明,集束化办法在处理不受监督的适应性个人再识别(ReID)任务方面行之有效,但是,在整个培训过程中,围绕这一办法的现有工作仍然受到杂音假标签和不可靠的概括化能力的影响。为解决这些问题,本文件提出一种新的方法,通过限制杂音假标签,以更概括化的能力来学习特征代表。首先,我们建议采用一种抽样退出方法,防止模型的培训落入经常使用杂音假标签的样品造成的恶性循环。此外,我们还在传统的相互教学结构下提出了一种全新的方法,称为“特征多样性学习(FDL)”,这可以大大提高目标领域特征代表的普遍化能力。实验结果表明,我们提议的FDL-SD在多个基准数据集上取得了最先进的业绩。