23rd January 2023

AnyFace: A Data-Centric Approach For Input-Agnostic Face Detection

ISSAI scientists Askat Kuzdeuov, Darina Koishigarina, and ISSAI Founding Director Professor Huseyin Atakan Varol have developed a new project called AnyFace. AnyFace is a data-centric approach to input-agnostic face detection. It is a simple but effective method to facilitate the labeling of different face datasets.

In computer vision, face detection is the first step in many face applications, such as face recognition, emotion recognition, age detection, virtual makeup, vital sign monitoring, and so on. Thanks to advances in deep learning for computer vision, numerous applications have been developed for human faces. 

Recently, other domains, such as animal species and cartoon characters, have also gained attention, but they are still far behind human faces in terms of the number of applications. The main challenge is the limited number of annotated face datasets in these domains. Annotating large-scale datasets is tedious and requires substantial human labor. In this regard, ISSAI presents an input-agnostic face detector, AnyFace, which facilitates the annotation of face datasets in various domains.

The model achieves accurate results in all domains—humans, animals and cartoon characters. It can be used to quickly and easily annotate face datasets in different domains, which will further facilitate the development of face applications.

For more information, please refer to our article:

Link to the demo page: 

Link to the source code: