Development and Evaluation of Automatic Facemask Template Generator

The COVID-19 pandemic has resulted in the normalization of wearing masks in public spaces. However, commercially available masks assume symmetrical faces and are unsuitable for faces deformed by diseases or accidents. Therefore, it is important to make face masks that fit the faces of individuals. To address this problem, we developed a prototype of an automatic facemask template generator for individuals using the three-dimensional surface data of their faces. In a previous study, we proposed the generation of a face mask template from a set of portrait and profile images of a user's face. This method was evaluated using airflow tests to check for air leakage around the masks. The results show that the proposed mask has a better protection ability; however, its fitness also affects the three-dimensional shapes of faces. To improve the fitting and droplet protection performance, we developed a new method for an automatic template generator using three-dimensional facial data. This study describes the basic architecture of the generator and the results of airflow and fitness tests as evaluations. Flow analysis was conducted using particle image velocimetry and a mannequin with three types of facemasks. The proposed mask conditionally reduced the flow rate in front of mouth compared to that in the other two masks. In the fitness experiments, two masks were fabricated using this software and the positional error for each feature point was measured. Subsequently, although the maximum error was 17.9 mm for the mental protuberance, the subjects subjectively felt a better fit with the proposed mask.