面向算力受限边缘环境的双分支多尺度感知人脸检测网络

doi:10.11959/j.issn.1000-436x.2020177

通信学报 ›› 2020, Vol. 41 ›› Issue (8): 165-174.doi: 10.11959/j.issn.1000-436x.2020177

面向算力受限边缘环境的双分支多尺度感知人脸检测网络

戚琦,马迎新,王敬宇(),孙海峰,廖建新

北京邮电大学网络与交换国家重点实验室，北京 100876

修回日期:2020-07-12 出版日期:2020-08-25 发布日期:2020-09-05
作者简介:戚琦（1982– ），女，河北廊坊人，博士，北京邮电大学副教授、博士生导师，主要研究方向为智能边缘计算、轻量级神经网络、业务网络智能化等|马迎新（1996- ），男，山西临汾人，北京邮电大学硕士生，主要研究方向为深度学习、计算机视觉、人脸检测与识别|王敬宇（1978- ），男，吉林长春人，博士，北京邮电大学教授、博士生导师，主要研究方向为智能网络、人工智能、计算机视觉、深度学习、多媒体通信等|孙海峰（1989– ），男，天津人，博士，北京邮电大学讲师、硕士生导师，主要研究方向为人工智能、机器视觉、自然语言处理、深度学习等|廖建新（1965– ），男，四川宜宾人，博士，北京邮电大学“长江学者”特聘教授、博士生导师，主要研究方向为移动通信网络、业务网络化、人工智能、多媒体业务等
基金资助:
国家重点研发计划基金资助项目(2018YFB1800502);国家自然科学基金资助项目(61671079);国家自然科学基金资助项目(61771068);北京市自然科学基金资助项目(4182041)

Multi-scale aware dual path network for face detection in resource-constrained edge computing environment

Qi QI,Yingxin MA,Jingyu WANG(),Haifeng SUN,Jianxin LIAO

State Key Laboratory of Networking and Switching Technology,Beijing University of Posts and Telecommunications,Beijing 100876,China

Revised:2020-07-12 Online:2020-08-25 Published:2020-09-05
Supported by:
The National Key Research and Development Program of China(2018YFB1800502);The National Natural Science Foundation of China(61671079);The National Natural Science Foundation of China(61771068);The Beijing Municipal Natural Science Foundation(4182041)

摘要/Abstract

摘要：

针对边缘算力受限，难以部署复杂结构的人脸检测深度神经网络的问题，为减少资源消耗，并保证人脸在多尺度变化、遮挡、模糊、光照等复杂场景下的检测精度，提出了多尺度感知的轻量化人脸检测算法。采用改进的人脸残差神经网络作为特征提取网络，并提出双分支浅层特征提取模块，并行分支理解图像多尺度信息，进而由深浅特征融合模块将底层图像信息与高层语义特征融合，配合多尺度感知的训练策略监督多分支学习差异化特征。实验结果表明，所提算法可有效提取多样化的特征，在保持模型高效性和低推理时延的同时，有效提升了算法的精度和稳健性。

关键词: 人脸检测, 多尺度感知, 特征融合, 人脸特征分析, 深度学习

Abstract:

Aiming at the problem that face detectors with complex deep neural structures are difficult to deploy in the resource-constrained edge computing environment,to reduce the resource consumption while maintain the accuracy in complex scenes such as multi-scale face changes,occlusion,blur,and illumination,SDPN(multi-scale aware dual path network) for face detection was proposed.The Face-ResNet (face residual neural network) was improved,and a dual path shallow feature extractor was used to understand the multi-scale information of the image through parallel branches.Then the deep and shallow feature fusion module,a combination of the underlying image information and the high-level semantic feature,was used in conjunction with the multi-scale awareness training strategy to supervise the multi-branch learning discriminating features.The experimental results show that SDPN can extract more diversified features,which effectively improve the accuracy and robustness of face detection while maintaining the efficiency of the model and low inference delay.

Key words: face detection, multi-scale aware, feature fusion, analysis of facial features, deep learning

中图分类号:

TP391.41

戚琦,马迎新,王敬宇,孙海峰,廖建新. 面向算力受限边缘环境的双分支多尺度感知人脸检测网络[J]. 通信学报, 2020, 41(8): 165-174.

Qi QI,Yingxin MA,Jingyu WANG,Haifeng SUN,Jianxin LIAO. Multi-scale aware dual path network for face detection in resource-constrained edge computing environment[J]. Journal on Communications, 2020, 41(8): 165-174.

图/表 10

图1

图2

图3

图4

表1

图5

表2

图6

图7

表3

参考文献 24

[1]	HU P , RAMANAN D . Finding tiny faces[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2017: 951-959.
[2]	CHI C , ZHANG S , XING J ,et al. Selective refinement network for high performance face detection[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Palo Alto:AAAI Press, 2019: 8231-8238.
[3]	LI J , WANG Y , WANG C ,et al. DSFD:dual shot face detector[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2019: 5060-5069.
[4]	NAJIBI M , SAMANGOUEI P , CHELLAPPA R ,et al. SSH:single stage headless face detector[C]// Proceedings of the IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2017: 4875-4884.
[5]	TANG X , DU D K , HE Z ,et al. Pyramidbox:a contextassisted single shot face detector[C]// Proceedings of the European Conference on Computer Vision. Berlin:Springer, 2018: 797-813.
[6]	HE Y , XU D , WU L ,et al. LFFd:a light and fast face detector for edge devices[J]. arXiv Preprint,arXiv:1904.10633, 2019
[7]	ZHANG S , ZHU X , LEI Z ,et al. Faceboxes:a CPU real-time face detector with high accuracy[C]// 2017 IEEE International Joint Conference on Biometrics. Piscataway:IEEE Press, 2017: 1-9.
[8]	YANG S , LUO P , LOY C C ,et al. Wider face:a face detection benchmark[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 5525-5533.
[9]	DENG J , DONG W , SOCHER R ,et al. ImageNet:a largescale hierarchical image database[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern. Piscataway:IEEE Press, 2009: 248-255.
[10]	REN S , HE K , GIRSHICK R ,et al. Faster R-CNN:towards real-time object detection with region proposal networks[C]// Neural Information Processing Systems. Massachusetts:MIT Press, 2015: 91-99.
[11]	LIN T Y , DOLLAR P , GIRSHICK R ,et al. Feature pyramid networks for object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern. Piscataway:IEEE Press, 2017: 2117-2125.
[12]	HUANG L , YANG Y , DENG Y ,et al. Densebox:unifying landmark localization with end to end object detection[J]. arXiv Preprint,arXiv:1509.04874, 2015
[13]	ZHANG K , ZHANG Z , LI Z ,et al. Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Processing Letters, 2016,23(10): 1499-1503.
[14]	JIANG H , LEARNED-MILLER E , . Face detection with the faster R-CNN[C]// 2017 12th IEEE International Conference on Automatic Face ＆ Gesture Recognition. Piscataway:IEEE Press, 2017: 650-657.
[15]	LIU W , ANGUELOV D , ERHAN D ,et al. SSD:single shot multibox detector[C]// Proceedings of the European Conference on Computer Vision. Berlin:Springer, 2016: 21-37.
[16]	REDMON J , DIVVALA S , GIRSHICK R ,et al. You only look once:unified,real-time object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 779-788.
[17]	ZHANG S , ZHU X , LEI Z ,et al. S3FD:single shot scaleinvariant face detector[C]// Proceedings of the IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2017: 192-201.
[18]	LAW H , DENG J . Cornernet:detecting objects as paired keypoints[C]// Proceedings of the European Conference on Computer Vision. Berlin:Springer, 2018: 734-750.
[19]	DUAN , KAIWEN , ,et al. Centernet:keypoint triplets for object detection[C]// Proceedings of the IEEE International Conference on Computer Vision. Piscataway:IEEE Press, 2019: 6569-6578.
[20]	LIU W , LIAO S , REN W ,et al. High-level semantic feature detection:a new perspective for pedestrian detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2019: 5187-5196.
[21]	HE K , ZHANG X , REN S ,et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2016: 770-778.
[22]	ODENA A , DUMOULIN V , OLAH C . Deconvolution and checkerboard artifacts[J]. Distill, 2016,1(10):e3.
[23]	SINGH B , DAVIS L S . An analysis of scale invariance in object detection SNIP[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway:IEEE Press, 2018: 3578-3587.
[24]	LIN T , GOYAL P , GIRSHICK R ,et al. Focal loss for dense object detection[C]// Proceedings of the IEEE International Conference on Computer vision. Piscataway:IEEE Press, 2017: 2980-2988.

实验	下采样卷积核	浅层残差结构	DPE提取模块	深浅特征融合		检测精度
实验	下采样卷积核	浅层残差结构	DPE提取模块	深浅特征融合	简单	中等	困难
1	3×3	×	×	×	0.927	0.922	0.673
2	4×4	×	×	×	0.933	0.925	0.674
3	5×5	×	×	×	0.934	0.925	0.677
4	4×4	√	×	×	0.940	0.928	0.767
5	4×4	√	√	×	0.940	0.931	0.792
6	4×4	√	√	√	0.937	0.927	0.849

模型		检测精度		推理速度/(frame.s^-1)	模型参数量/MB
模型	简单	中等	困难	推理速度/(frame.s^-1)	模型参数量/MB
FTF^[1]	0.819	0.797	0.638	5.3	114.54
DSFD^[3]	0.949	0.936	0.845	1.7	457.99
PyramidBox^[5]	0.945	0.934	0.850	3.8	256.61
SSH^[4]	0.919	0.907	0.814	3.8	75.41
SDPN	0.937	0.927	0.849	6.7	65.06

模型		检测精度		时延/ms	速度/(frame.s^-1)
模型	简单	中等	困难	时延/ms	速度/(frame.s^-1)
SDPN	0.937	0.927	0.849	55.8	17.9
LFFD^[6]	0.896	0.865	0.770	31.27	31.98
FaceBoxes3.2x^[7]	0.791	0.794	0.715	25.37	39.41

面向算力受限边缘环境的双分支多尺度感知人脸检测网络

Multi-scale aware dual path network for face detection in resource-constrained edge computing environment

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 24

相关文章 15

Metrics

推荐阅读 0

[1]	陈东昱, 陈华, 范丽敏, 付一方, 王舰. 基于深度学习的随机性检验策略研究[J]. 通信学报, 2023, 44(6): 23-33.
[2]	李荣鹏, 汪丙炎, 张宏纲, 赵志峰. 知识增强的语义通信接收端设计[J]. 通信学报, 2023, 44(6): 70-76.
[3]	马帅, 裴科, 祁华艳, 李航, 曹雯, 王洪梅, 熊海良, 李世银. 基于生成模型的地磁室内高精度定位算法研究[J]. 通信学报, 2023, 44(6): 211-222.
[4]	杨洁, 董标, 付雪, 王禹, 桂冠. 基于轻量化分布式学习的自动调制分类方法[J]. 通信学报, 2022, 43(7): 134-142.
[5]	杨秀璋, 彭国军, 李子川, 吕杨琦, 刘思德, 李晨光. 基于Bert和BiLSTM-CRF的APT攻击实体识别及对齐研究[J]. 通信学报, 2022, 43(6): 58-70.
[6]	廖勇, 王世义. 高速移动环境下基于RM-Net的大规模MIMO CSI反馈算法[J]. 通信学报, 2022, 43(5): 166-176.
[7]	廖育荣, 王海宁, 林存宝, 李阳, 方宇强, 倪淑燕. 基于深度学习的光学遥感图像目标检测研究进展[J]. 通信学报, 2022, 43(5): 190-203.
[8]	赵增华, 童跃凡, 崔佳洋. 基于域自适应的Wi-Fi指纹设备无关室内定位模型[J]. 通信学报, 2022, 43(4): 143-153.
[9]	廖勇, 程港, 李玉杰. 基于深度展开的大规模MIMO系统CSI反馈算法[J]. 通信学报, 2022, 43(12): 77-88.
[10]	段雪源, 付钰, 王坤, 李彬. 基于简单统计特征的LDoS攻击检测方法[J]. 通信学报, 2022, 43(11): 53-64.
[11]	霍俊彦, 邱瑞鹏, 马彦卓, 杨付正. 基于最邻近帧质量增强的视频编码参考帧列表优化算法[J]. 通信学报, 2022, 43(11): 136-147.
[12]	康海燕, 冀源蕊. 基于本地化差分隐私的联邦学习方法研究[J]. 通信学报, 2022, 43(10): 94-105.
[13]	刘威, 陈成, 江锐, 卢涛. 四通道无监督学习图像去雾网络[J]. 通信学报, 2022, 43(10): 210-222.
[14]	张红霞, 王琪, 王登岳, 王奔. 基于深度学习的区块链蜜罐陷阱合约检测[J]. 通信学报, 2022, 43(1): 194-202.
[15]	晏燕, 丛一鸣, Adnan Mahmood, 盛权政. 基于深度学习的位置大数据统计发布与隐私保护方法[J]. 通信学报, 2022, 43(1): 203-216.