Facial point annotations


Existing facial databases cover large variations including: different subjects, poses, illumination, occlusions etc. However, the provided annotations appear to have several limitations.




 Figure 1:  (a)-(d) Annotated images from MultiPIE, XM2VTS, AR, FRGC Ver.2 databases, and (e) examples from XM2VTS with inaccurate annotations.


  1. The majority of existing databases provide annotations for a relatively small subset of the overall images.
  2. The accuracy of provided annotations in some cases is not so good (probably due to human fatigue). 
  3. The annotation model of each database consists of different number of landmarks. 

These problems make cross-database experiments and comparisons between different methods almost infeasible. To overcome these difficulties, we propose a semi-automatic annotation methodology for annotating massive face datasets. This is the first attempt to create a tool suitable for annotating massive facial databases. 

All the annotations are provided for research purposes ONLY (NO commercial products).



 Figure 2: The 68 points mark-up used for our annotations.



We employed our tool for creating annotations (following the Multi-PIE 68 points mark-up, please see Fig. 2) for the following databases:



Please cite as:


Christos Sagonas - c.sagonas@imperial.ac.uk / Stefanos Zafeiriou - s.zafeiriou@imperial.ac.uk