Masked Deep Face Recognition using ArcFace and Ensemble Learning

With advancements in technology, human biometrics, especially face recognition, has witnessed a tremendous increase in usage, prominently in the field of security. Face recognition proves to be a convenient, coherent, and efficient way to identify a person uniquely. Face recognition systems are trained generally on human faces sans masks. With the ubiquitous use of face masks due to the ongoing COVID-19 pandemic, face recognition becomes a daunting challenge. In this paper, the deep learning architectures, namely MobileNetV2, DenseNet201, ResNet50V2, and VGG16 with the ArcFace loss function, were trained on the newly created dataset called "MaFaR", which consists of a mixture of masked and unmasked images of 75 distinct individuals, and ensemble learning techniques have been used to improve the performance, achieving an accuracy 93.65%. © 2021 IEEE.

Masked Face Detection Using Artificial Intelligent Techniques

Nowadays, wearing a face mask is a vital routine in life, but threats are increasing in public due to the advantage of wearing face masks. Existing works do not perfectly detect the human face and also not possible to apply for different faces detection. To overwhelm this issue, in this paper we proposed real-time face mask detection. The proposed work consists of six steps: video acquisition and keyframes selection, data augmentation, facial parts segmentation, pixel-based feature extraction, Bag of Visual Words (BoVW) generation, and face mask detection. In the first step, a set of keyframes are selected using the histogram of gradient (HoG) algorithm. Secondly, data augmentation is involved with three steps as color normalization, illumination correction (parameterized CLAHE), and pose normalization (Angular Affine Transformation). In the third step, facial parts are segmented using the clustering approach i.e., Expectation Maximization with Gaussian Mixture Model (EM-GMM), in which facial regions are segmented into Eyes, Nose, Mouth, Chin, and Forehead. Then, CapsNet based Feature Extraction is performed using CapsNet approach, which performance is higher and lightweight model than the Yolo Tiny V2 and Yolo Tiny V3, and extracted features are constructed into Codebook by Hassanat Similarity with K-Nearest neighbor (H-M with KNN) algorithm. For mask detection, L2 distance function is used. Experiments conducted using Python IDLE 3.8 for the proposed model and also previous works as GMM with Deep learning (GMM + DL), Convolutional Neural Network (CNN) with VGGF, Yolo Tiny V2, and Yolo Tiny V3 in terms of various performance metrics. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.