Kinect SDK 1.5 - Face Tracking : WPF color-depth only tracker

Question

I'm working on the new kinect SDK for face tracking and I noticed some differences between managed and unmanaged code provided with the examples (Previous Question). I can't reproduce the fast c++ example using c# and WPF: starting with the same setting with kinect for windows, here's the two sets of code:

c++:

if (m_KinectSensorPresent && m_KinectSensor.GetVideoBuffer())
{
    HRESULT hrCopy = m_KinectSensor.GetVideoBuffer()->CopyTo(m_colorImage, NULL, 0, 0);
    if (SUCCEEDED(hrCopy) && m_KinectSensor.GetDepthBuffer())
    {
        hrCopy = m_KinectSensor.GetDepthBuffer()->CopyTo(m_depthImage, NULL, 0, 0);
    }
    // Do face tracking
    if (SUCCEEDED(hrCopy))
    {
        FT_SENSOR_DATA sensorData(m_colorImage, m_depthImage, m_KinectSensor.GetZoomFactor(), m_KinectSensor.GetViewOffSet());

        FT_VECTOR3D* hint = NULL;
        if (SUCCEEDED(m_KinectSensor.GetClosestHint(m_hint3D)))
        {
            hint = m_hint3D;
        }
        if (m_LastTrackSucceeded)
        {
            hrFT = m_pFaceTracker->ContinueTracking(&sensorData, hint, m_pFTResult);
        }
        else
        {
            hrFT = m_pFaceTracker->StartTracking(&sensorData, NULL, hint, m_pFTResult);
        }
    }
}

c#:

int hr;
            HeadPoints headPointsObj = null;
            Vector3DF[] headPoints = GetHeadPointsFromSkeleton(skeletonOfInterest);

        if (headPoints != null && headPoints.Length == 2)
        {
            headPointsObj = new HeadPoints { Points = headPoints };
        }

        this.copyStopwatch.Start();
        this.colorFaceTrackingImage.CopyFrom(colorImage);
        this.depthFaceTrackingImage.CopyFrom(depthImage);
        this.copyStopwatch.Stop();

        var sensorData = new SensorData(this.colorFaceTrackingImage, this.depthFaceTrackingImage, DefaultZoomFactor, Point.Empty);
        FaceTrackingSensorData faceTrackSensorData = sensorData.FaceTrackingSensorData;

        this.startOrContinueTrackingStopwatch.Start();
        if (this.trackSucceeded)
        {
            hr = this.faceTrackerInteropPtr.ContinueTracking(ref faceTrackSensorData, headPointsObj, this.frame.ResultPtr);
        }
        else
        {
            hr = this.faceTrackerInteropPtr.StartTracking(
                ref faceTrackSensorData, ref regionOfInterest, headPointsObj, this.frame.ResultPtr);
        }

The result of the same tracking call, providing the same data apparently, result false for c# and true for c++.

Any idea? I want my tracker to work besides the skeleton data, like the c++ example.

MDaldoss MDaldoss · Accepted Answer · 2012-05-31T14:17:24

I found a solution, provided in this forum Microsoft Forum. You have to modify the tools the SDK gives to you, and it's not nice to mess with provided code, but it works, finally.

Kinect SDK 1.5 - Face Tracking : WPF color-depth only tracker

1 Answers