Names and Faces Demonstrator v1.0

The following interface allows to browse the images and texts of the YAHOO! database and to visualize the output results of several algorithms :
  • The Oxford's face recognition algorithm.
  • The KUL's entities analyzer algorithm.
  • The INRIA's "Names To Faces" clustering algorithm.
The interface is composed of 6 frames :
  • Left frame allows to browse the directories of the YAHOO! database and to select one specific image.
  • Upper central frame displayes the selected image. On the image, each face recognized by the Oxford's face recognition algorithm is represented by a yellow square (* see next paragraph).
  • Lower central frame displayes the text associated with the image. Each word recognized and associated with a person by the KUL's entity recognizer is represented in blue (* see next paragraph).
  • Upper right frame displayes all text entities of type person (* see next paragraph) recognized by the KUL's entity recognizer algorithm.
  • Middle right frame displayes the ground truth names (* see next paragraph) associated with each detected face.
  • Lower right frame displayes the names (* see next paragraph) associated with each recognized face by the INRIA's clustering algorithm.
The interface offers interactivity :
  • Moving the mouse over an element marked with an '*' in the preceding paragraph will highlight in a different colour all others associated elements.
  • Clicking on a name in the middle right frame (Ground Truth Name) will display a popup window that will allows to change the Ground Truth Name value.
Note :
  • The specific value "_UNKNOWN_" is used to indicate a name which has no value.
  • The specific value "< none >" indicates an empty list.

Select KUL's text entities version :

Contact: Yves Gufflet (engineer),, phone +33 4 7661 5383
Bill Triggs (coordinator),, phone +33 4 7661 5233
GRAVIR-INRIA, 655 avenue de l'Europe, Montbonnot 38330, France