arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2601.01537cs.CV
收藏

FAR-AMTN:面向人脸属性识别的注意力多任务网络

FAR-AMTN: Attention Multi-Task Network for Face Attribute Recognition

School of Electronic and Information Engineering, Tongji University(同济大学电子与信息工程学院);School of Electronics and Computer Science, University of Southampton(南安普顿大学电子与计算机科学学院)

Gong Gao, Zekai Wang, Xianhui Liu, Weidong Zhao

AI总结:

FAR-AMTN通过引入加权共享组特定注意力模块和跨组特征融合机制,提升了人脸属性识别的泛化性能和准确性,同时显著减少了模型参数量。

摘要:

为了提高多任务网络(MTN)在人脸属性识别(FAR)中的泛化性能,有必要有效共享多个相关预测任务之间的相关信息。传统MTN方法创建共享低级模块和不同的高级模块,导致随着任务增加模型参数呈指数级增长。这种做法也限制了高级特征之间的交互,阻碍了属性之间语义关系的探索,从而负面影响泛化能力。为此,本研究引入了FAR-AMTN,一种新的面向FAR的注意力多任务网络。它集成了具有共享参数的加权共享组特定注意力(WSGSA)模块,以最小化复杂性并提高组特征表示。此外,利用跨组特征融合(CGFF)模块促进属性组之间的交互,增强特征学习。还引入了动态加权策略(DWS)以实现任务同步收敛。在CelebA和LFWA数据集上的实验表明,所提出的FAR-AMTN在参数显著更少的情况下表现出更高的准确性。

英文摘要:

To enhance the generalization performance of Multi-Task Networks (MTN) in Face Attribute Recognition (FAR), it is crucial to share relevant information across multiple related prediction tasks effectively. Traditional MTN methods create shared low-level modules and distinct high-level modules, causing an exponential increase in model parameters with the addition of tasks. This approach also limits feature interaction at the high level, hindering the exploration of semantic relations among attributes, thereby affecting generalization negatively. In response, this study introduces FAR-AMTN, a novel Attention Multi-Task Network for FAR. It incorporates a Weight-Shared Group-Specific Attention (WSGSA) module with shared parameters to minimize complexity while improving group feature representation. Furthermore, a Cross-Group Feature Fusion (CGFF) module is utilized to foster interactions between attribute groups, enhancing feature learning. A Dynamic Weighting Strategy (DWS) is also introduced for synchronized task convergence. Experiments on the CelebA and LFWA datasets demonstrate that the proposed FAR-AMTN demonstrates superior accuracy with significantly fewer parameters compared to existing models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑