Double layer multiple task learning for age estimation with insufficient training samples

Double layer multiple task learning for age estimation with insufficient training samples
复制标题

训练样本不足的情况下的双层多任务学习年龄估计

DOI:
10.1016/j.neucom.2014.06.047
复制
发表时间:
2015-01
期刊:
影响因子:
6
通讯作者:
Nanning Zheng
Nanning Zheng
中科院分区:
计算机科学2区
文献类型:
--
作者:
Jianyi Liu;Xi Yang;Yuehu Liu;Nanning Zheng

文献摘要

参考文献

被引文献

相似文献

缺乏训练样本是人脸年龄估计的主要难点之一。本文指出,将年龄估计作为多任务学习(MTL)问题来处理,可以减轻训练样本问题的影响。在此基础上,我们利用多类分数函数重新构造了年龄估计任务,并提出了一种双层多任务学习(DLMTL)方法。在主题层,采用个性化的年龄估计模型和全局模型,实现不同主题间共同衰老模式知识的共享;在年龄标签层,进一步对任意特定年龄标签上的分数函数估计子任务进行建模,充分利用沿年龄轴的序列信息。所提出的DLMTL模型可以表述为非常简洁的内积表示,最后利用多核学习(multiple kernel learning, MKL)工具进行求解。在FG-NET和MORPH老化数据库上的实验结果验证了我们的方法优于许多其他流行的年龄估计算法,特别是在训练样本极度不足的应用中。
One of the main difficulty of facial age estimation is the lack of training sample problem. In this paper, we point out that when age estimation is treated as a multiple task learning (MTL) problem, the impact of training sample problem can be relieved. By this idea, we re-formulate the age estimation task using the multi-class score function and develop a double layer multiple task learning (DLMTL) approach. In the subject layer, the personalized age estimation models as well as the global model are used to share knowledge of common aging pattern among different subjects; in the age label layer, the sub-tasks of score function estimation on any specific age label are further modeled to fully exploit the sequential information along the age axis. The proposed DLMTL model can be formulated into a very concise inner product representation, and it is finally solved using the multiple kernel learning (MKL) tool. The experimental results upon the FG-NET and MORPH aging databases verified that our method outperforms many other popular age estimation algorithms especially for the extremely training sample insufficient applications.
DOI: 10.1016/j.neucom.2012.05.005
发表时间: 2012-11
期刊: Neurocomputing
影响因子: 6
作者:
Zeng-Shun Zhao;Li Zhang;Meng Zhao;Z. Hou;Changshui Zhang
通讯作者: Zeng-Shun Zhao;Li Zhang;Meng Zhao;Z. Hou;Changshui Zhang
DOI: 10.1109/icme.2007.4284917
发表时间: 2007-07
期刊: 2007 IEEE International Conference on Multimedia and Expo
影响因子: --
作者:
Yun Fu;Ye Xu;Thomas S. Huang
通讯作者: Yun Fu;Ye Xu;Thomas S. Huang
DOI: --
发表时间: 2010
期刊: --
影响因子: --
作者:
通讯作者: --
DOI: 10.1016/j.sigpro.2013.07.025
发表时间: 2014
期刊: Signal Process.
影响因子: --
作者:
Jianyi Liu;Yao Ma;Lixin Duan;Fangfang Wang;Yuehu Liu
通讯作者: Jianyi Liu;Yao Ma;Lixin Duan;Fangfang Wang;Yuehu Liu
DOI: 10.1109/iccv.2011.6126454
发表时间: 2011-11
期刊: 2011 International Conference on Computer Vision
影响因子: --
作者:
Jie Luo;T. Tommasi;B. Caputo
通讯作者: Jie Luo;T. Tommasi;B. Caputo