ATF: An Alternating Training Framework for Weakly Supervised Face Alignment

doi:10.1109/TMM.2022.3164798

CORC > 自动化研究所 > 中国科学院自动化研究所 > 复杂系统认知与决策实验室

	ATF: An Alternating Training Framework for Weakly Supervised Face Alignment
	Lan, Xing 1,2; Hu, Qinghao 2; Cheng, Jian 1,2
刊名	IEEE TRANSACTIONS ON MULTIMEDIA
	2023
卷号	25 页码:1798-1809
关键词	Face alignment multi-task learning weakly supervised
ISSN号	1520-9210
DOI	10.1109/TMM.2022.3164798
通讯作者	Cheng, Jian(jcheng@nlpr.ia.ac.cn)
英文摘要	In recent years, various face-landmark datasets have been published. Intuitively, it is significant to integrate multiple labeled datasets to achieve higher performance. Due to the different annotation schemes of datasets, it is hard to directly train models using them together. Although numerous efforts have been made in the joint use of datasets, there remain three shortages in previous methods, i.e., additional computation, limitation of the markups scheme, and limited support for the regression method. To solve the above issues, we proposed a novel Alternating Training Framework (ATF), which leverages the similarity and diversity across multiple datasets for a more robust detector. ATF mainly contains two sub-modules: Alternating Training with Decreasing Proportions (ATDP) and Mixed Branch Loss (L-MB). In particular, ATDP trains multiple datasets simultaneously via a weakly supervised way to take advantage of the diversity among them, and L-MB utilizes similar landmark pairs to constrain different branches of the corresponding datasets. Besides, we extend the framework to easily handle three situations: single target detector, joint detector, and novel detector. Extensive experiments demonstrate the effectiveness of our framework for both heatmap-based and direct coordinate regression. Moreover, we have achieved a joint detector that outperforms state-of-the-art methods on each benchmark.
资助项目	National Key Research and Development Program of China[2021ZD0201504] ; National Natural Science Foundation of China[62106267] ; Jiangsu Key Research and Development Plan[BE2021012-2] ; Jiangsu Leading Technology Basic Research Project[BK20192004]
WOS关键词	NETWORK
WOS研究方向	Computer Science ; Telecommunications
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:001007432100016
资助机构	National Key Research and Development Program of China ; National Natural Science Foundation of China ; Jiangsu Key Research and Development Plan ; Jiangsu Leading Technology Basic Research Project
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/53699]
专题	复杂系统认知与决策实验室
通讯作者	Cheng, Jian
作者单位	1.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 2.Chinese Acad Sci CASIA, Natl Lab Pattern Recognit NLPR, Inst Automat, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Lan, Xing,Hu, Qinghao,Cheng, Jian. ATF: An Alternating Training Framework for Weakly Supervised Face Alignment[J]. IEEE TRANSACTIONS ON MULTIMEDIA,2023,25:1798-1809.
APA	Lan, Xing,Hu, Qinghao,&Cheng, Jian.(2023).ATF: An Alternating Training Framework for Weakly Supervised Face Alignment.IEEE TRANSACTIONS ON MULTIMEDIA,25,1798-1809.
MLA	Lan, Xing,et al."ATF: An Alternating Training Framework for Weakly Supervised Face Alignment".IEEE TRANSACTIONS ON MULTIMEDIA 25(2023):1798-1809.