DSPNet: A low computational-cost network for human pose estimation

doi:10.1016/j.neucom.2020.11.003

	DSPNet: A low computational-cost network for human pose estimation
	Zhong, Fujin 2; Li, Mingyang 2; Zhang, Kun 1; Hu, Jun 2; Liu, Li 2
刊名	NEUROCOMPUTING
	2021-01-29
卷号	423 页码:327-335
关键词	Human pose estimation Convolutional neural network Up-sampling unit Deep supervision Multi-scale obtaining
ISSN号	0925-2312
DOI	10.1016/j.neucom.2020.11.003
英文摘要	Existing human pose estimation methods usually have a high computational load, which is very unfavorable for resource-limited equipment. To address this issue, we propose a low computational-cost deep supervision pyramid network called DSPNet. Firstly, we design a lightweight up-sampling unit instead of transposed convolution as a decoder for the network. In the case of decreased computation, it has brought an increase in prediction accuracy. Secondly, we present a novel deep supervision pyramid architecture to improve the multi-scale obtaining ability of MSRA SimpleBaseline while not bringing any increase in the number of parameters. The experimental results on both MPII and COCO pose estimation benchmarks illustrate that DSPNet achieves almost equivalent state-of-the-art results with a low computational load. Especially, the computational cost of DSPNet is 12.7% of SimpleBaseline and the estimation accuracy is improved by 0.9 points when both methods use the same backbone network (EfficientNet) on MPII validation set. The code of the proposed method is availabe at https://github.com/sumaliqinghua/ DSPNet. (C) 2020 Elsevier B.V. All rights reserved.
资助项目	National Key Research and Development Program of China[2017YFC0804002] ; National Natural Science Foundation of China[61876027] ; National Natural Science Foundation of China[61751312] ; Chongqing Research Program of Basic Research and Frontier Technology[cstc2017jcyjAX0406] ; National Natural Science Foundation of Chongqing[cstc2019jcyjcxttX0002]
WOS研究方向	Computer Science
语种	英语
出版者	ELSEVIER
WOS记录号	WOS:000599909500010
内容类型	期刊论文
源URL	[http://119.78.100.204/handle/2XEOYT63/16555]
专题	中国科学院计算技术研究所
通讯作者	Zhong, Fujin
作者单位	1.Chinese Acad Sci, Inst Comp Technol, Beijing, Peoples R China 2.Chongqing Univ Posts & Telecommun, Chongqing Key Lab Computat Intelligence, Chongqing, Peoples R China
推荐引用方式 GB/T 7714	Zhong, Fujin,Li, Mingyang,Zhang, Kun,et al. DSPNet: A low computational-cost network for human pose estimation[J]. NEUROCOMPUTING,2021,423:327-335.
APA	Zhong, Fujin,Li, Mingyang,Zhang, Kun,Hu, Jun,&Liu, Li.(2021).DSPNet: A low computational-cost network for human pose estimation.NEUROCOMPUTING,423,327-335.
MLA	Zhong, Fujin,et al."DSPNet: A low computational-cost network for human pose estimation".NEUROCOMPUTING 423(2021):327-335.