A quite comprehensive survey of facial gender recognition methods in computer vision, this paper may even help some to feel reassured in these uncertain times.

The paper starts by describing the main application fields of facial gender recognition. Beyond security and surveillance, coming tragically into the foreground lately, facial gender recognition is also important in fields such as biometrics and human-computer interaction. The main problems and challenges facial gender recognition poses are investigated next, among which are those connected with image capture of any object, such as lighting and illumination or image resolution, and those connected more intimately with recognition of a human face, such as age, ethnicity, or facial expression (smiling, neutral, angry, and so on); additional elements such as facial hair, glasses, hats, or other forms of clothing can paradoxically both hinder and help at the same time. Then, the most important techniques are reviewed, such as fiducial distances, pixel intensity values, rectangle features, local binary patterns, scale-invariant feature transforms, and Gabor wavelets; this is the central and most technical part of the paper. Each technique is briefly described, and the original methods and studies behind it are referenced. Finally, a performance survey of these techniques is also given, not directly but by citing test runs performed against the most common datasets available (basically, catalogs of faces).

As explained in the title of the paper, the authors’ goal is not to propose any new facial recognition method, but rather to review and compare existing ones. In this, their mission is fully accomplished--even people completely outside the field can be very satisfactorily introduced to it with this paper.