Face tracking

Return to Tracking Configuration

Introduction

Unifeye face tracking enables you to detect a face in the input image/video stream, follow its movements and track it in six degrees of freedom.

For the face tracking to work properly, lighting conditions are important. The face should be well lit with the lighting coming from the front and being homogeneous. If the use case allows, the autofocus of the camera should be turned off.

Creating a face tracking configuration file

Open any editor of your choice and create a TrackingData file containing the following:

  • A FaceSensorSource has to be added to the Sensors section

    • SensorID can be freely chosen

    • UsedMethod must be set to CAD

    • RotateImage has to be set to one of the following three values:

      • -1 if the camera delivering the input has been rotated 90 degrees counter clock wise

      • 0 if the camera delivering the input is used in normal landscape mode

      • 1 if the camera delivering the input has been rotated 90 degrees clock wise

    • MinFaceSize can be set in two different ways:

      • If a value between 0 and 1 is set, this number is treated as the factor of the original image size that a potential face structure has to have to be treated as a candidate for a face. (If a value of 0.5 is set, a potential face has to cover at least half of the input image, a value of 0.33 means that a potential face has to take up at least a third of the input image)

      • If a value greater than 1 is set, it defines the minimum number of pixels a potential face has to take up in the input image.

    • DetermineGender defines whether or not a gender detection takes place. (Note: Currently this tag is not used!)

      • 0 : no gender detection

      • 1 : gender detection is turned on

    • PD can be set to the pupil distance (in mm) of the target person(s). This parameter is optional and can be used to make the camera-pose more accurate.

    • SensorCosID can be freely chosen

  • A COS has to be added to the Connections section

    • Using an AverageFuser helps to reduce jittering when tracking faces

      • SingleCOSHistoryFusion is set to 1 if the single sensor cos history fusion shall be activated and 0 otherwise

      • SingleCOSHistoryFusionNumber specifies the number of entries in the pose history that shall be averaged

    • A SensorSource has to be added to the COS

      • The SensorID has to be the same as defined above in the Sensors section

      • The SensorCosID has to be the same as defined above in the Sensors section

      • The default COS position is right between the eyes of the tracked face. If the COS is supposed to be located somewhere else, the COSOffset values have to be set.

A valid face tracking configuration file would look like this:

Using face tracking configuration files in Unifeye

To activate face tracking within the Unifeye SDK you simply have to load an according tracking configuration file and use the API function setTrackingData() (see http://doxygen.metaio.com/UnifeyeSDK/UnifeyeSDK_doxygen/group__Configuration.html). You can also load the according tracking configuration file within the Unifeye GUI using the Tracking Configuration Tool:

mapGeneratorLoadingTrackingConfiguration.png

Note: There is an example 3D extensible tracking configuration file called AS_Tracking_Data_FaceTracking.xml in <UnifeyeInstallationFolder>/config/SensorSources.


Return to Tracking Configuration


-- SupportMetaio - 2011-01-21

Topic attachments
I Attachment Action Size Date Who Comment
pngpng faceTrackingTD.png manage 39.5 K 2011-01-26 - 16:22 SupportMetaio  
Topic revision: r5 - 2011-02-10 - 18:26:38 - SupportMetaio
 
This site is powered by the TWiki collaboration platformCopyright &© by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback