VisionWorks object tracker vs VPI KLT tracker

I have been playing around with the new VPI demos. The KLT tracker seems to do very similar things to the VisionWorks object tracker…

I realise VisionWorks is on the way out, but which tracker is more accurate and reliable? I realise this is a bit of a subjective question but I’m hoping someone knows the answer.



Both VisionWorks and VPI are based on the same KLT algorithm.
So the accuracy should be identical.

  1. Simon Baker, Iain Matthews, “Lucas-Kanade 20 Years On: A Unified Framework”.
    International Journal of Computer Vision, February 2004, Volume 56, issue 3, pp 221-255.


But VisionWorks also uses the FAST algorithm. Does the VPI tracker uses this? I see no mention of it, so I am wondering if VisionWorks is better/faster in this example. From the VisionWorks documentation:

Tracker uses the FAST algorithm to find corners and passes them through the Optical Flow to the next frame. Corners it finds are filtered by motion distance, direction and Optical Flow precision to remove outliers. With the rest corners, named “key points” below, the algorithm tries to map each key point with key points tracked from the previous frames to set the weights of the new key points. Then using these key points for each object, the bounding box center is found, and then the bounding box scale is estimated.

FAST Algorithm

Features from accelerated segment test (FAST) is a corner detection method originally developed by Edward Rosten and Tom Drummond. For more information, see:


No, but FAST has been requested before:

In general, we still recommend VPI since VisionWorks won’t be updated anymore.


1 Like