基于边缘计算的人脸识别模型研究

Research on Face Recognition Model Based on Edge Computing

随着对人脸识别系统要求的不断提高,为提升用户使用体验、提高识别效率、降低安全风险,借鉴边缘计算实现思路,采用MTCNN检测和MobileNet网络相结合的方式构建人脸识别模型。MTCNN具有资源要求低、精度较高等特点,而MobileNet网络基于深度可分离思路降低了网络模型参数,通过对损失函数和对比验证相似度参数进行改进,实现了在资源受限设备中的快速高效的人脸识别功能。

With the increasing requirements for face recognition system,in order to improve the user experience,improve the recognition efficiency and reduce security risks,the edge computing approach is used to construct a face recognition model combining MTCNN detection and MobileNet network.MTCNN has the characteristics of low resource requirements and high accuracy,while MobileNet network reduces the network model parameters based on the idea of deep separability.By improving the loss function and the similarity parameter of comparative verification,it realizes fast and efficient face recognition function in resource constrained devices.