Before you read further, try out the on-line version of the "wheel" face navigation interface at http://www.ece.ubc.ca/~tzupei/facenav.htm. It's fun!

Conventional face navigation systems focus on finding new faces via facial features. This method, though intuitive, has limitations. Notably, it is geared towards distinctive features, hence, does not work effectively for finding typical faces. We investigate an alternative approach to search and navigate through an overall face configuration space. For this, we implemented an interface which shows gradients of faces arranged spatially using an n-dimensional norm-based face generation method. It is like a colour-wheel but with faces (as shown in the image below).

Because our interface allows users to observe faces holistically, the facial composition information is not lost during searching, as compared to a face component method. We compared our gradient based face navigation system with a typical static slider-based system in a navigation task as well as comparing with a hybrid dynamic slider system. (The two images below illustrate the appearance of our wheel interface and slider interface.)

Results from our first study show that our method is more effective in allowing users to concentrate on face navigation when compared with a static slider interface. This is helpful in a face matching task as it reduces the number of times users re-examine faces. Results from our second study further indicate that our interface is more effective in coping with correlated navigation axes when compared with a dynamic slider interface. Our third and fourth study confirmed that while slider-based interface is more suited for converging to close proximity to the target face, gradient-based interface caters for matching refinement. While it may be counter-intuitive that sliders, which work so well as an interface for colour navigation, are inadequate for face matching tasks, our results suggest that new interfaces, such as our gradient-based system and dynamic sliders, are necessary for navigation in higher dimensional face space.

Downloads

Related Links

The Singular Inversions homepage

On-line face navigation Game

Contact Information