Face Recognition With Masks Based on Spatial Fine-Grained Frequency Domain Broadening

Along with social distancing, wearing masks is an effective method of preventing the transmission COVID-19 in ongoing pandemic. However, occlude a large number facial features, recognition. The recognition rate existing methods may be significantly reduced by presence masks. In this paper, we propose to effectively solve problem lack feature information needed perform on people proposed approach uses image super-resolution technology preprocessing along deep bilinear module improve EfficientNet. It also combines enhancement frequency domain broadening, fuses spatial features and unoccluded areas face, classifies fused features. area are increased accuracy masked faces. results cross-validation show that achieved 98% RMFRD dataset, as well higher faster speed than previous methods. addition, performed experimental evaluation actual system 99%, which demonstrates effectiveness practicability method.