In the past three years, the entire world has been exposed to a new virus, COVID-19. It has become very difficult to deal with the fingerprint system, as the exchange of things and contact with devices by more than one person is a good way to transmit the virus. Accordingly, many institutions and organizations have resorted to using face prints as an alternative to handprints to identify people and record the attendance of employees in general. Since the virus is a virus that infects the respiratory system, people are forced to wear face masks or the so-called masks to avoid transmission of the virus during coughing by the infected person. Since the wearing of masks greatly affects the facial features, it was necessary to come up with a technology that allows the identification of masked faces, and this is the subject of our study. Since the face loses many of its features while wearing masks, an algorithm was proposed to recognize the face and train the convolutional neural networks (CNN) through some of the main facial features, including 3D imaging, as it was concluded that the triple loss does not apply to our data sets, as 3D selection has less loss compared to 2D image, Due to its ability to select all feature samples from feature spaces with larger distances between layers and reduced distances between regions, a large cosine loss was utilized as the training loss function. In order to reach a model that deals more with areas not covered by the mask, the input unit was designed, as its function is to combine the Inception- Resent unit and the Convolutional Mass unit, and so, any portion of the face that isn't covered gains weight, thus increasing the importance of those areas in the recognition of masked faces, and experiments that were conducted on several sets of masked faces data showed that the algorithm works to significantly increase the accuracy of masking face recognition, and it can accurately recognize the face using masks.