Difference between revisions of "Attention-system-vision"

From Wiki for iCub and Friends
Jump to navigation Jump to search
 
 
(2 intermediate revisions by the same user not shown)
Line 1: Line 1:
The goal of this group is to track the face of the person in front of the iCub. This module is already implemented by [http://eris.liralab.it/wiki/VVV12_Participants#Ugo_Pattacini Ugo]. Here is a summery of the steps needed to run this demo.
The goal of this group is to identify the face of two persons in front of the iCub.
 
=== Demo of face tracking with the iCub ===
 
As a starting point, we found out that there is a demo in yarp that is detecting and tracking the face of one human. This demo was implemented by [http://eris.liralab.it/wiki/VVV12_Participants#Ugo_Pattacini Ugo]. Here is a summery of the steps needed to run this demo.
 
=== Goals ===
 
  Ideally, we would like to '''detect''' (and not directly track) '''several''' faces, so that a specific saliency map can be created and inserted in the visual attention system.
 
=== Using openCV to detect faces (or eyes) : ===
 
  It seems to be possible to detect multiple faces in an image with openCV, using haar transforms, as explained here :
  http://www.scribd.com/doc/4547752/opencv-objectdetection-2007june10
  http://www.cognotics.com/opencv/servo_2007_series/part_2/sidebar.html
 
  There seems to be some example code there :
  http://robotworkshop-kumotek.blogspot.it/2011/04/programming-opencv22-face-detection.html
  If not using CV_HAAR_FIND_BIGGEST_OBJECT", multiple results (faces) might be returned.
 
  Eyes might possibly be detected too :
  http://python.dzone.com/articles/face-and-eyes-detection-opencv

Latest revision as of 16:06, 24 July 2012

The goal of this group is to identify the face of two persons in front of the iCub.

Demo of face tracking with the iCub

As a starting point, we found out that there is a demo in yarp that is detecting and tracking the face of one human. This demo was implemented by Ugo. Here is a summery of the steps needed to run this demo.

Goals

 Ideally, we would like to detect (and not directly track) several faces, so that a specific saliency map can be created and inserted in the visual attention system.

Using openCV to detect faces (or eyes) :

 It seems to be possible to detect multiple faces in an image with openCV, using haar transforms, as explained here :
 http://www.scribd.com/doc/4547752/opencv-objectdetection-2007june10
 http://www.cognotics.com/opencv/servo_2007_series/part_2/sidebar.html
 
 There seems to be some example code there :
 http://robotworkshop-kumotek.blogspot.it/2011/04/programming-opencv22-face-detection.html
 If not using CV_HAAR_FIND_BIGGEST_OBJECT", multiple results (faces) might be returned.
 
 Eyes might possibly be detected too :
 http://python.dzone.com/articles/face-and-eyes-detection-opencv