CelebA+masks is a dataset of synthetic face images wearing masks. The method of synthesis in this dataset is to create synthetic masks and then superimpose them on the faces of the original images. The dataset involves 9 masks of different colors, shapes, and fabrics. This method relies on Spark AR Studio, a program developed by Facebook for creating Instagram face filters. Researchers used this method to CASIA-WebFace dataset445,446 (90%) mask samples were generated for CelebA dataset196,254 (96.8%) mask samples were generated.

Due to the spread of the COVID-19 pandemic, people are wearing masks, which poses a challenge to traditional face recognition. This dataset is used to study face recognition when the mouth and nose are blocked.