In [1], we utilized MCT-based face detection on the Cerf/FIFA eye-tracking data set to model the influence of faces on visual saliency. Most importantly, we evaluated different integration strategies (linear, sub-linear, supra-linear, and quaternion) in combination with quaternion DCT image signature saliency and investigated the importance of scaling, rotation and shape of an (elliptical) Gaussian face model (the previous work by Cerf et al. only considered circular models).

To give a short impression of the face detections:


Related Work

[1] B. Schauerte, G. A. Fink, "Predicting Human Gaze using Quaternion DCT Image Signature Saliency and Face Detection". In Proceedings of the 12th IEEE Workshop on the Applications of Computer Vision (WACV), Breckenridge, CO, USA, January 9-11, 2012.

Download: [pdf] [bibtex] [code #1]


You can download the face detections (in a text format) either individually (direct) or in a compressed .zip file. Furthermore, we provide a Matlab script (.m) to generate and visualize exemplary face saliency maps (be aware that you additionally have to scale w0 and h0 appropriately, if you want scaled models to achieve optimal performance).

   [1] Face Detections: [direct] [.zip]
   [2] Matlab Visualization Script: [.m]