3D face-based Automatic recognition of facial attributes (expressions, age, gender, ethnicity) – 3D Face Analyzer

Human face conveys a significant amount of information, including information about head orientation, the identity, emotional state, gender, age, ethnic origin, education level, etc., and plays important role in face-to-face communication between human. The use of these facial clues during interaction is made possible by the remarkable human ability to recognize and interpret faces and facial behaviors. The present project aims at automatic interpretation of 3D face images so that contactless human-computer interaction based on typical user's facial attributes, such as facial expressions, gender, age and ethnic origin, can be developed for an improved HCI.

While most of the face attribute sensitive applications are still unfathomable, especially in the field of human-computer interaction, one can already figure out some simple direct applications, including for instance reliable face recognition, user’s affect recognition based on facial expression, age and gender specific human computer interaction, etc.

From a scientific perspective, machine-based recognition of facial attributes is also a challenging problem for multiple research communities, including machine learning, computer vision, neurosciences, etc. For instance, age estimation from facial images is not a classic learning problem as typically the aging progress is unpredictable over the time and different people age in different way while temporal learning data are usually very rare. Indeed, apart from biological factors, external factors as diverse as ethnicity, climatic conditions, food intake, mental stress, etc. also contribute towards aging effects, it is thus natural to expect different individuals to age differently. For all these reasons, automatic interpretation of face images has attracted increasing interest from several research communities in the recent years. However, most of these works are mainly based on the analysis of 2D texture face images although studies have shown that facial attributes such as the aging, gender or ethnicity are not only revealed by the 2D textures, but also has close relationship with the 3D morphology of the human faces.

The project 3D Face Analyzer proposes to join the effort of two French teams (ECL Liris and USTL LIFL) and two Chinese teams (Beihang Irip and NCUT IIP) having each significant expertise in face processing and targets reliable recognition of facial attributes on 2.5D or 3D face models, thus making use of face 3D morphology, texture and landmarks at the same time. While developing 3D analysis-based techniques directly aiming at recognition of facial attributes, we also want to make forward knowledge on some underlying fundamental issues, e.g. stability of discrete geometric measures and descriptions (curvature, distance, etc.) across variations in terms of model resolution and precision, 3D non-rigid surface registration and matching in the presence of noisy data. Another important aim of the project is the identification and collection of significantly representative resources of 3D face models in facial expressions, age and gender for the purpose of training and testing.