Human Face Recognition Dataset (HFRD) with the Impacts of COVID-19 Pandemic (v2020)

The HFRD (v2020) repository includes 4835 images of masked human faces that are valuable for training and testing various automatic face recognition methods. The proposed HFRD dataset is developed in this paper based on the impact of the COVID-19 pandemic’s stage, which requires wearing additional Personal Protective Equipment (PPE), such as facial masks. Thus, new standard images are introduced to be applied in the field of the Partial Facial Recognition (PFR) domain. Therefore, this enhanced HFRD dataset is also useful for different Artificial Intelligence (AI) tools, particularly, in the decision-support system as it attempts to provide further knowledge for predicting the spread of COVID-19. The HFRD repository is mainly composed of three subsets, which include; MaskedMUCT, MaskedFASSEG, and MaskedAT&T datasets that address various orientations of faces, conditions of illumination, and many other facial expressions. First, the MaskedMUCT subset contains 3,755 coloured images that provide a diversity of lighting, age, and ethnicity, which are divided into five subset folders, each containing 751 images. Second, the MaskedFASSEG dataset contains 680 coloured images of masked facial images, which are categorised into four folders as follows: MaskedFASSEG-frontal01, MaskedFASSEG-frontal02, MaskedFASSEG-frontal03, and MaskedFASSEG-multipose01, and which include 70, 70, 150, and 390 masked RGB facial images, respectively. The last subset-folder, namely, ‘MaskedAT&T’, consists of 400 mono-coloured masked faces that are constructed to represent the frontal face of segment partitions. For the entire face images, many kinds of PPE-based facial masks are added to cover the nose, mouth, and chin area, and what remains visible from the face includes eyes, forehead, and hair