How to 'train' the detectobject node?

Like the title says: how can I create the necessary xml training file, in order to detect the object I want it to detect. (in my case, I would like to know position, size and rotation around the Z-axis of a white rectangular piece of paper).

hei ft,

this is more an opencv question really since vvvv is merely using opencvs DetectObject function here. i remember there being a separate file called haartraining.exe that came with the opencv installation. maybe you can find some information here: http://opencv.willowgarage.com/documentation/object_detection.html

hi , i,ve been interested in knowing thsi for quite a long time ,ft let us know if you find the way please .

regards

@colorsound: so far no luck. Haven’t been able to put a lot of time in it yet…

would be nice to have the predator tracking within vvvv
http://info.ee.surrey.ac.uk/Personal/Z.Kalal/tld.html
i think it went opensource too https://github.com/zk00006/OpenTLD

I know people were working on a c version too, because the original was written in matlab I think. I also suppose it won’t take too long before TLD will be a part of OpenCV, which would be ideal, since OpenCv plugins already exist.

I don’t know if TLD knows about rotation though, since it ‘learns’ different appearances of the same object.