MASKED FACE RECOGNITION BY ZEROING THE MASKED REGION WITHOUT MODEL RETRAINING

With the recent global pandemic event, the requirement to use masks, espe-cially in public spaces, has become a challenge to the existing face recognition system. To overcome this challenge, previous studies have performed transfer learning and finetuning of the existing model with masked face datasets. Others have performed a preprocessing by cropping the masked face and then fine-tuning the model with the newly cropped datasets. However, retraining with preprocessed or masked faces may be costly or even unavailable for some with limited resources. Furthermore, these methods of preprocessing are ill-advised to be used directly using models that are not retrained as was found in this study. Therefore, this study explores and presents a way of cropping which shows increases in performance without the requirement of any training to the existing face recognition model. This method managed to increase the performance of the existing model by up to 9.09% when presented with masked-face scenarios.