This example uses the segmenter plugin with the selfie segmentation model as a matte for a compact Dual Kawase blur pipeline. A few small offscreen passes build the blur efficiently, then the final pass keeps the segmented foreground sharp over the mirrored webcam feed.

It is a good reference for multi-pass webcam compositing when you want a cleaner person outline than pose masks usually give, without paying for a huge one-pass kernel.