OpenCV background substraction with prior data

I like to extract a foreground image with one of the OpenCV background subtraction methods (BackgroundSubtractorKNN,BackgroundSubtractorMOG2 or BackgroundSubtractorGMG) with sometimes complicated conditions.

However, I have a priori data for a part of the image, describing the propability of each pixel to classify as foregorund or background.

Is it possible with any of this algorithms above to feed my prior data for increased quality? Or maybe any pre / post operation that would make use of it and also affect the background training?