论文阅读 [TPAMI-2023] Consistency and Diversity Induced Human Motion Segmentation

论文阅读 [TPAMI-2023] Consistency and Diversity Induced Human Motion Segmentation

论文搜索(studyai.com)

搜索论文: Consistency and Diversity Induced Human Motion Segmentation

搜索论文: http://www.studyai.com/search/whole-site/?q=Consistency+and+Diversity+Induced+Human+Motion+Segmentation&fr=csdn

关键字(Keywords)

Motion segmentation; Computer vision; Task analysis; Transfer learning; Data models; Clustering algorithms; Correlation; Subspace clustering; human motion segmentation; transfer learning; multi-level representation

机器学习; 机器视觉; 运筹与优化

迁移学习; 聚类; 子空间聚类; 运动捕捉; ADMM

摘要(Abstract)

Subspace clustering is a classical technique that has been widely used for human motion segmentation and other related tasks.

子空间聚类是一种经典的技术，已广泛用于人体运动分割和其他相关任务。

However, existing segmentation methods often cluster data without guidance from prior knowledge, resulting in unsatisfactory segmentation results.

然而，现有的分割方法常常在没有先验知识指导的情况下对数据进行聚类，导致分割结果不理想。

To this end, we propose a novel Consistency and Diversity induced human Motion Segmentation (CDMS) algorithm.

为此，我们提出了一种新的一致性和多样性诱导的人体运动分割（CDMS）算法。

Specifically, our model factorizes the source and target data into distinct multi-layer feature spaces, in which transfer subspace learning is conducted on different layers to capture multi-level information.

具体来说，我们的模型将源数据和目标数据分解为不同的多层特征空间，其中在不同的层上进行转移子空间学习以捕获多层信息。

A multi-mutual consistency learning strategy is carried out to reduce the domain gap between the source and target data.

为了减少源数据和目标数据之间的领域差距，实施了一种多相互一致性学习策略。

In this way, the domain-specific knowledge and domain-invariant properties can be explored simultaneously.

通过这种方式，可以同时探索领域特定知识和领域不可变财产。

Besides, a novel constraint based on the Hilbert Schmidt Independence Criterion (HSIC) is introduced to ensure the diversity of multi-level subspace representations, which enables the complementarity of multi-level representations to be explored to boost the transfer learning performance.

此外，引入了一种基于希尔伯特-施密特独立准则（HSIC）的新约束，以确保多级子空间表示的多样性，这使得可以探索多级表示的互补性，以提高转移学习性能。

Moreover, to preserve the temporal correlations, an enhanced graph regularizer is imposed on the learned representation coefficients and the multi-level representations of the source data.

此外，为了保持时间相关性，在学习的表示系数和源数据的多级表示上施加了增强的图正则化器。

The proposed model can be efficiently solved using the Alternating Direction Method of Multipliers (ADMM) algorithm.

使用交替方向乘子法（ADMM）算法可以有效地求解所提出的模型。

Extensive experimental results on public human motion datasets demonstrate the effectiveness of our method against several state-of-the-art approaches…

在公共人体运动数据集上的大量实验结果证明了我们的方法对几种最先进方法的有效性。

作者(Authors)

[‘Tao Zhou’, ‘Huazhu Fu’, ‘Chen Gong’, ‘Ling Shao’, ‘Fatih Porikli’, ‘Haibin Ling’, ‘Jianbing Shen’]