Masked Faces Recognition Using Transfer Learning

Through this study, we present our study that aims to implement a masked face recognition model by using a transfer learning algorithm for facial recognition and masked faces datasets. To achieve the goal, we first implemented an algorithm built on a Facenet model. Our choices come down to the considerable Facenet's outcomes model When training on the well-known LFW database. As a result, he gave an accuracy rate that exceeds 98%. About masked faces, we used two different datasets that are Simulated Masked Faces Recognition Dataset (SMFRD) and Real world Masked Faces Dataset (RWMFD). Then we have established a performance comparison between the two datasets. Thereby, we tried to adjust our model to be robust against partial occlusions caused by wearing the mask. Relative to other state-of-the-art models, obtained outcomes on the Real-World-Masked-Face-Dataset display high recognition achievement.