Comparison of various landmark detecting techniques in the context of forensic facial recognition

In this paper two different landmark detection algorithms (Dlib and STASM) have been compared to each other. First a quantitative study has been done where the landmarks were used for ˙Secondly, a qualitative study has been done in which the location of the landmarks detected by Dlib and STASM were compared with landmarks placed by human examiners. STASM detected a face on a location where there was none in 7,6% of all images, because of this Dlib got better results in the quantitative study in almost every situation in comparison with STASM. In the qualitative study the human examiners placed their landmarks closer to the Dlib landmarks than to the STASM landmarks. So the Dlib landmarks are more precise than the STASM landmarks according to the human examiners.