Multi-Task Learning with Knowledge Transfer for Facial Attribute Classification

机译：具有知识转移的多任务学习，用于面部属性分类

获取原文

获取原文并翻译 | 示例

页面导航

摘要
著录项
相似文献
相关主题

摘要

Having achieved satisfying performance in multiple areas, multi-task learning (MTL) is being applied on facial attribute classification. However, most multi-task learning algorithms neglect the latent connections among facial attributes, sorting attributes based on local information only, or merely viewing each attribute as independent. The concept of curriculum learning suggests that in multi-task learning, "easy" tasks can be learned first and used to guide the learning process of "hard" tasks. Inspired, we propose KT-MTL, a novel MTL network with knowledge transfer for facial attribute classification. Depending only on label information, attributes are divided into multiple tasks by spectral clustering and labeled as "strong" or "weak" embodying their correlation extent. During training, parameters learned in "strong" network are transferred to "weak" net, imitating the teacher-student learning process. Both parts contribute to the total loss with a specifically designed loss function. The proposed network archives a competitive overall accuracy score of above 92% on aligned CelebA images and the highest accuracy of 91.89% on "weak" tasks.

机译：在多个领域都取得了令人满意的性能后，多任务学习（MTL）应用于面部属性分类。然而，大多数多任务学习算法忽略了面部属性之间的潜在联系，仅基于本地信息对属性进行排序，或者仅将每个属性视为独立的。课程学习的概念表明，在多任务学习中，可以首先学习“简单”任务，并用来指导“困难”任务的学习过程。受此启发，我们提出了KT-MTL，这是一种新颖的MTL网络，该网络具有用于面部属性分类的知识转移。仅根据标签信息，通过频谱聚类将属性划分为多个任务，并标记为“强”或“弱”，体现它们的相关程度。在训练过程中，在“强”网络中学习到的参数将传输到“弱”网络中，从而模仿了师生的学习过程。这两个部分都通过专门设计的损耗函数来贡献总损耗。拟议的网络在对齐的CelebA图像上归档的竞争性总体准确性得分超过92％，在“弱”任务上的最高准确性达到91.89％。

著录项

来源
《2019 IEEE International Conference on Industrial Technology》|2019年|877-882|共6页
会议地点 Melbourne(AU)
作者
Xiaohui Fanhe; Jie Guo; Zheng Huang; Weidong Qiu; Yuele Zhang;
展开▼
作者单位

Department of Cyberspace Security, Shanghai Jiao Tong University, Shanghai, China;

Department of Cyberspace Security, Shanghai Jiao Tong University, Shanghai, China;

Department of Cyberspace Security, Shanghai Jiao Tong University, Shanghai, China;

Department of Cyberspace Security, Shanghai Jiao Tong University, Shanghai, China;

Department of Cyberspace Security, Shanghai Jiao Tong University, Shanghai, China;

展开▼
会议组织
原文格式 PDF
正文语种 eng
中图分类
关键词
Task analysis; Facial features; Correlation; Training; Knowledge transfer; Face; Computer architecture;

机译：任务分析;面部特征;相关性;训练;知识转移;面孔;计算机体系结构;;
入库时间 2022-08-26 14:32:42

相似文献

外文文献
中文文献
专利

1. Dermoscopic attributes classification using deep learning and multi-task learning [J] . Irek Saitov, Tatyana Polevaya, Andrey Filchenkov Procedia Computer Science . 2020,第5期

机译：利用深度学习和多任务学习的Dermoscopic属性分类
2. A Multi-Task Framework for Facial Attributes Classification through End-to-End Face Parsing and Deep Convolutional Neural Networks [J] . Nature reviews Cancer . 2020,第2期

机译：通过端到端脸部解析和深卷积神经网络分类的面部属性的多任务框架
3. Island Loss for Improving the Classification of Facial Attributes with Transfer Learning on Deep Convolutional Neural Network [J] . Shuvendu Roy International Journal of Image, Graphics and Signal Processing . 2020,第1期

机译：深卷积神经网络转移学习改善面部属性分类的岛屿损失
4. Multi-Task Learning with Knowledge Transfer for Facial Attribute Classification [C] . Xiaohui Fanhe, Jie Guo, Zheng Huang, International Conference on Industrial Technology . 2019

机译：具有面部属性分类的知识转移的多任务学习
5. Bayesian multi-task learning for clustering and classification with non-parametric priors. [D] . An, Qi. 2008

机译：贝叶斯多任务学习，用于使用非参数先验进行聚类和分类。
6. A Multi-Task Framework for Facial Attributes Classification through End-to-End Face Parsing and Deep Convolutional Neural Networks [O] . Khalil Khan, Muhammad Attique, Rehan Ullah Khan, 2020

机译：通过端到端脸部分析和深度卷积神经网络进行面部属性分类的多任务框架
7. A Multi-Task Framework for Facial Attributes Classification through End-to-End Face Parsing and Deep Convolutional Neural Networks [O] . Khalil Khan, Muhammad Attique, Rehan Ullah Khan, 2020

机译：通过端到端脸部解析和深卷积神经网络分类的面部属性的多任务框架

Multi-Task Learning with Knowledge Transfer for Facial Attribute Classification

摘要

著录项

相似文献

相关主题

期刊订阅