CIFaceDetector is a face detector for Ikaros that uses Apple's Core Image Face Detector. OS X Only.

The module finds all faces in an image as well as the eyes and the mouth. Additional inputs tells whether the face is smiling and if it is novel. Positions and sizes are given as fractions of the image size.

Example XML definition

A simple example

      class = "CIFaceDetector"
      name = "CIFaceDetector"


classThe class name for the module; must be "CIFaceDetector".string
nameThe name of this instance of the module.string
max_facesmaximum number of facesint10
min_sizeminimum size of a face (0.2-1.0)float0.1
use_trackingtrack faces between framesboolyes
detect_smilesdetect smilesboolyes
detect_blinksdetect eye blinkboolyes
mouth_correctionmap the mouth position from the detector onto a line that is perpendicular to the line between the eyes and centered between the eyes; this mostly gives a better position than the original detector. boolyes

Module Connections


INPUTThe input image


FACE_POSITIONLocation of each face in the image
FACE_SIZESize of each face (width, height)
FACE_BOUNDSFace bounding boxes in image (x0, y0,...x3, y3)
EYE_LEFT_POSITIONLocation of the left eyes in the image (0-1)
EYE_RIGHT_POSITIONLocation of the right eyes in the image (0-1)
MOUTH_POSITIONLocation of the right eyes in the image (0-1)
ROTATIONThe head tilt in degrees
SMILEThis output is 1 if the corresponding face is smiling; 0 otherwise
BLINK_LEFTThis output is 1 if the left eye is blinking; 0 otherwise
BLINK_RIGHTThis output is 1 if the right eye is blinking; 0 otherwise
NOVELTYThis output is 1 if the face just appeared in the image
OBJECT_IDA unique ID for the face. New IDs are assigend everytime a face dissapears and reappears
LIFEThis output increases with 1 for each tick that the face is tracked
FACE_COUNTThe number of detected faces in the image


OS X only.


Christian Balkenius
Lund Univeristy Cognitive Science



blog comments powered by Disqus