This work presents a novel approach to face recognition that recognizes faces both with and without facial masks worn during the Covid-19 pandemic. The two-fold contribution of this study are: evaluation of the hand-crafted descriptors for facial mask detection and deep learning based method for face recognition in the presence of a mask. In this study, Local Binary Pattern (LBP), Histogram of Oriented Gradients (HOG), Local Directional Order Pattern (LDOP), have been used along with Linear Support Vector Machine (SVM) for facial mask detection and has proven to result in a very high success rate of 99.6029 %. This study has classified the Face Recognition system into two models: subject dependent (or closed-set) and subject independent (or open-set). Conventional solutions to face recognition use the entire facial image as the input to their algorithms. This thesis presents an approach where hand-crafted feature descriptors are used to identify the presence of a mask, If detected, switch to ocular recognition for those subjects wearing mask else face recognition is performed. The ensemble method is incorporated which uses two CNNs trained with face and ocular region of the identities where the system dynamically switches between the CNNs when the mask is detected in the input image. Open-set and closed-set approaches followed in experiments are conducted on the Real-World Masked Face Recognition Dataset (RMFRD) the database of masked and unmasked faces. Finally, the analysis shows how the granularity of this experiment can be leveraged to obtain an improved accuracy rate over the traditional full-face recognition approach in the presence of a face covering.