We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
云边协同的轻量级网络结构 人脸识别方法.
- Authors
祁春阳; 黄杰; 赵翔宇; 汪周红
- Abstract
To solve the problem that face recognition cannot be deployed in Internet of Things(IoT)edge devices with limited computing and storage resources due to the use of convolutional neural networks with high complexity, a lightweight Mobile multi-mask convolutional neural network(MTCNN)face detection and recognition algorithm that can be deployed on IoT edge devices using a cloud-edge collaboration technology is proposed to reduce the resource overhead of IoT edge devices. In the face detection algorithm, a lightweight sub-network is used to obtain the position of the face frame and the coordinates of the face key points, and a IMobileNet network model based on the Mobile-Inception and Mobile-Resent structures is designed in the face recognition algorithm. Experimental results show that when the accuracy of the face detection algorithm is reduced by 1% on average compared with the existing algorithm, although the missed detection rate increases by 1% on average, the running time is reduced by more than 20% on average; when the face recognition algorithm achieves the same accuracy as the MobileNet V2 network, the average recognition time of the designed and selected IMobileNet-Small network is reduced by more than 12%. Therefore, the lightweight face detection and recognition algorithms can maintain recognition accuracy and meet the requirements of the edge device deployment.
- Publication
Journal of Southeast University / Dongnan Daxue Xuebao, 2023, Vol 53, Issue 1, p1
- ISSN
1001-0505
- Publication type
Article
- DOI
10.3969/j.issn.1001-0505.2023.01.001