Evolutionary Algorithms for Face Recognition with Mask

Abstract

Due to the Covid-19 pandemic, wearing masks in public places has become a necessity. But it also comes with its challenges, existing face recognition systems are trained to recognize faces with all the features and therefore are failing to work efficiently due to masks. To provide a potential solution to this problem and to recognize faces with masks two evolutionary algorithms, Crow Search Algorithm (CSA) and Cuttle Fish Algorithm (CFA), are used for feature selection which select an optimal subset of features from the existing dataset with vast number of features. In the last step four machine learning classifiers (Support Vector Machine, Random Forest classifier, K-Nearest Neighbor, and Decision tree classifier) are practiced on each subset of features received by both the feature selection algorithms. Experimental results show that CSA removed most of the irrelevant features by selecting only 41% of the original featured and CFA selected 60% of the features. Highest accuracy of classification was received by CSA of 86.5% with Random Forest classifier. Therefore, it shows that CSA and CFA can be used in various other real time applications due to their reduced computational cost and high accuracy.