Mask usage recognition using vision transformer with transfer learning and data augmentation

The COVID-19 pandemic has disrupted various levels of society. use masks is essential in preventing the spread by identifying an image a person using mask. Although only 23.1% people correctly, Artificial Neural Networks (ANN) can help classify good to slow Covid-19 virus. However, it requires large dataset train ANN that correctly. MaskedFace-Net suitable consisting 137016 digital images with 4 class labels, namely Mask, Mask Chin, Mouth and Nose Mouth. classification training utilizes Vision Transformers (ViT) architecture transfer learning method pre-trained weights on ImageNet-21k, random augmentation. In addition, hyper-parameters 20 epochs, Stochastic Gradient Descent (SGD) optimizer rate 0.03, batch size 64, Gaussian Cumulative Distribution (GeLU) activation function, Cross-Entropy loss function are used be applied three architectures ViT, Base-16, Large-16, Huge-14. Furthermore, comparisons without augmentation conducted. This study found best ViT Using this dataset, research reaches accuracy 0.9601 data, 0.9412 validation 0.9534 test data. shows model data improves mask usage, even better than convolutional-based Residual Network (ResNet).