Masked Frontal-Faces Database
Examples of generated masked face images
Masked Frontal-Faces Database (MFFD) is intended for use (modeling, training, and testing) in face recognition systems in which the position of frontal-faces is not only motivated, but also necessary for a formally correct interactive with the system.
Currently, MFFD is built on 100 original face images from the CUHK Face Sketch Database (CUFS). Based on these 100 originals, more than 20,000 images of masked faces of various types, shapes, textures, and colors were generated. Additionally, to the generated faces images in masks, their populations are added, including changing the color of the face skin, the color of the masks and the background, changing the texture (by smoothing and applying noise) and changing the proportions of faces as part of de-identification procedures, as well as arbitrary changes in the original size of the face images in masks while maintaining their proportions (affine transformations). All generated masked face images are saved in two formats – «PNG» and «GIF». The total amount of memory required to store MFFD is over 1.5 GB.
To download MFFD, you should fill in the form.
A detailed description of the generating face images method for the MFFD corpus is presented in:
Kukharev G.A., Ryumina E.V., Shulgin N.A. Method for generating masks on face images and systems for their recognition // Scientific and Technical Journal of Information Technologies, Mechanics and Optics, 2022, vol. 22, no. 3, pp. 547–558. DOI: 10.17586/2226-1494-2022-22-3-547-558.