Parameters for the wallflower dataset for mixture of gaussian algorithm

Question

I am trying to implement the MOG algorithm which is the standard algorithm in openCV for background modeling.

Right now I am a beginner to this so I am experimenting with the wallflower algorithm dataset by microsoft research and I cant figure out the parameters properly. I am just not satisfied with the O/P. Lots of false +ve/-ve.

I am working on the waving tree's sub part of the dataset.http://research.microsoft.com/en-us/um/people/jckrumm/WallFlower/TestImages.htm

Could some one suggest me the parameters for this dataset.

Answer 1

I think I got a good enough decent output with 5 gaussians, 2 VT, 0.99 background ratio and 0.15 learning rate.