Facial Mask Detection Using Depthwise Separable Convolutional Neural Network Model During COVID-19 Pandemic

• DOI: 10.3389/fpubh.2022.855254
• Publication Date: 2022-03-07T07:17:52Z
• Authors: Muhammad Zubair A...; Fahad R. Albogamy; Mabrook S. Al-Rak...; Junaid Asghar; Mohd Khairil Rahmat; Muhammad Mansoor ...; Adidah Lajis; Haidawati Mohamad...
• Keywords: Convolution (computer science); Benchmark (surveying); Contextual image classification

Deep neural networks have made tremendous strides in the categorization of facial photos last several years. Due to complexity features, enormous size picture/frame, and severe inhomogeneity image data, efficient face classification using deep convolutional remains a challenge. Therefore, as data volumes continue grow, effective mobile context utilizing advanced learning techniques is becoming increasingly important. In recent past, some Learning (DL) approaches for identify images been designed; many them use (CNNs). To address problem mask recognition images, we propose Depthwise Separable Convolution Neural Network based on MobileNet (DWS-based MobileNet). The proposed network utilizes depth-wise separable convolution layers instead 2D layers. With limited datasets, DWS-based performs exceptionally well. decreases number trainable parameters while enhancing performance by adopting lightweight network. Our technique outperformed existing state art when tested benchmark datasets. When compared Full baseline methods, results this study reveal that Convolution-based significantly improves (Acc. = 93.14, Pre. 92, recall F -score 92).