Description of Invention
Masked face recognition presents unique challenges, especially due to occlusion caused by face masks. However, the Histogram of Oriented Gradients (HOG) has been widely recognised for its effectiveness in capturing essential facial features while deep transfer learning which takes advantage of pre-trained models on large-scale datasets always can achieve optimal outcomes on specific tasks like masked face recognition. Therefore, this research method utilises the visualisation of HOG images as input to the EfficientNetV2-S pre-trained model for deep transfer learning to address the challenge of recognising masked face images.