基于改进的YOLOv3口罩佩戴检测和识别

计算机工程与科学 ›› 2022, Vol. 44 ›› Issue (10): 1812-1821.

基于改进的YOLOv3口罩佩戴检测和识别

任小康，刘行行

（西北师范大学计算机科学与工程学院，甘肃兰州 730070）

收稿日期:2021-03-08 修回日期:2021-06-03 接受日期:2022-10-25 出版日期:2022-10-25 发布日期:2022-10-28
基金资助:
国家自然科学基金（61662070,61762079）

Mask wearing detection and recognition based on the improved YOLOv3

REN Xiao-kang,LIU Xing-xing

（School of Computer Science and Engineering,Northwest Normal University，Lanzhou 730070，China）

Received:2021-03-08 Revised:2021-06-03 Accepted:2022-10-25 Online:2022-10-25 Published:2022-10-28

摘要/Abstract

摘要： 新冠疫情仍在全球肆虐，佩戴口罩可以有效阻断新冠病毒传播，口罩佩戴检测系统能及时提醒公共场所活动的人佩戴口罩。针对该问题及小尺度目标检测困难的问题，提出了一种基于YOLOv3改进的网络模型Face_mask Net用于口罩佩戴检测。由于YOLOv3算法训练的网络模型对小目标检测率低,IoU值相同时不能反映预测框和目标框是否相交，以及传统NMS对于遮挡经常产生错误抑制情况，Face_mask Net改进了残差块和神经网络结构，引入SPP模块和CSPNet网络模块，并采用DIoU作为损失函数,DIoU-NMS算法作为分类器。实验结果表明，Face_mask Net可以有效提高目标检测准确率，AP75下的平均准确率为58.05%，相比由YOLOv3算法训练的网络模型提高了4.11%。

关键词: YOLOv3, DIoU, SPP, 口罩佩戴检测, CSPNet

Abstract: The COVID-19 epidemic is still rampant around the world. Wearing masks can effectively block the spread of novel coronavirus, while mask wearing detection can timely remind people in public places to wear masks. To solve the problem and the difficulty of small scale target detection, an improved network model Face_mask Net based on the YOLOv3 algorithm is proposed for mask wearing detection. Because the network model trained by the YOLOv3 algorithm has a low detection rate of small targets,the same IoU value cannot reflect whether the prediction frame and the target frame intersect, and the traditional NMS often produces false suppression for occlusion, the algorithm in this paper improves the residual block and neural network structure, introduces SPP module and CSPNet network module, and adopt DIoU as the loss function and DIoU-NMS as the classifier. The experimental results show that Face_mask Net can effectively improve the target detection accuracy, and the average accuracy of AP75 is 58.05%, which is 4.11 percentage points higher than that of the network model trained by the Yolov3 algorithm.

Key words: YOLOv3, DIoU, spatial pyramid pooling(SPP), mask wearing detection, CSPNet

任小康, 刘行行. 基于改进的YOLOv3口罩佩戴检测和识别[J]. 计算机工程与科学, 2022, 44(10): 1812-1821.

REN Xiao-kang, LIU Xing-xing. Mask wearing detection and recognition based on the improved YOLOv3[J]. Computer Engineering & Science, 2022, 44(10): 1812-1821.

参考文献［19］

［1］	Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation［C］∥Proc of 2014 IEEE Conference on Computer Vision and Pattern Recognition,2014: 580-587.
［2］	Girshick R.Fast R-CNN［C］∥Proc of the IEEE International Conference on Computer Vision,2015: 1440-1448.
［3］	He K M,Zhang X Y,Ren S Q,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(9): 1904-1916.
［4］	Ren S Q,He K M,Girshick R,et al.Faster R-CNN:Towards real-time object detection with region proposal networks［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,39(6): 1137-1149.
［5］	Redmon J,Divvala S,Girshick R,et al.You only look once: Unified,real-time object detection［C］∥Proc of 2016 IEEE Conference on Computer Vision and Pattern Recognition,2016: 779-788.
［6］	Redmon J,Farhadi A.YOLO9000: Better,faster,stronger［C］∥Proc of 2017 IEEE Conference on Computer Vision and Pattern Recognition,2017: 7263-7271.
［7］	Redmon J,Farhadi A.YOLOv3: An incremental improvement［J］.arXiv:1804.02767,2018.
［8］	Johnson J W.Automatic nucleus segmentation with Mask-RCNN［C］∥Proc of the Science and Information Conference,2019: 399-407.
［9］	Wang C Y,Liao H Y M,Wu Y H,et al.CSPNet: A new backbone that can enhance learning capability of CNN［C］∥Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2020: 390-391.
［10］	Zheng Z H,Wang P, Liu W,et al.Distance-IoU loss: Faster and better learning for bounding box regression［C］∥Proc of the AAAI Conference on Artificial Intelligence,2020: 12993-13000.
［11］	Rahman M A,Wang Y. Optimizing intersection-over-union in deep neural networks for image segmentation［C］∥Proc of the International Symposium on Visual Computing,2016: 234-244.
［12］	Neubeck A, van Gool L.Efficient non-maximum suppression［C］∥Proc of the 18th International Conference on Pattern Recognition,2006: 850-855.
［13］	Rezatofighi H,Tsoi N,Gwak J Y,et al.Generalized intersection over union: A metric and a loss for bounding box regression［C］∥Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2019: 658-666.
［14］	Muhammad U,Wang W Q, Chattha S P,et al.Pre-trained VGGNet architecture for remote-sensing image scene classification［C］∥Proc of the 24th International Conference on Pattern Recognition,2018: 1622-1627.
［15］	He K M, Zhang X Y, Ren S Q, et al.Deep residual learning for image recognition［C］∥Proc of 2016 IEEE Confe- rence on Computer Vision and Pattern Recognition,2016: 770-778.
［16］	Lin T Y, Dollár P,Girshick R,et al.Feature pyramid networks for object detection［C］∥Proc of 2017 IEEE Confe- rence on Computer Vision and Pattern Recognition,2017: 2117-2125.
［17］	Shuo Y, Luo P, Loy C-C,et al.WIDER FACE: A face detection benchmark［C］∥Proc of 2016 IEEE Conference on Computer Vision and Pattern Recognition,2016: 5525-5533.
［18］	Ge S M, Li J, Ye Q T,et al.Detecting masked faces in the wild with lle-cnns［C］∥Proc of 2017 IEEE Conference on Computer Vision and Pattern Recognition,2017: 2682-2690.
［19］	Wang Z Y, Wang G C, Huang B J,et al.Masked face recognition dataset and application［J］. arXiv:2003.09093,2020.