Masked Face Recognition System Based on Attention Mechanism

With the continuous development of deep learning, face recognition field has also developed rapidly. However, with massive popularity COVID-19, masks is a problem that now about to be tackled in practice. In recognizing wearing mask, mask obscures most facial features face, resulting general model only capturing part information. Therefore, existing models are usually ineffective faces masks. This article addresses this and proposes an improvement Facenet. We use ConvNeXt-T as backbone network add ECA (Efficient Channel Attention) mechanism. enhances feature extraction unobscured obtain more useful information, while avoiding dimensionality reduction not increasing complexity. design new by investigating effects different attention mechanisms on data set ratios experimental results. addition, we construct large so can efficiently quickly train model. Through experiments, our proved 99.76% accurate for real A combined accuracy 99.48% extreme environments such too high or lousy contrast brightness.