When enabled, DeepHoldout only outputs an alpha matte of the main input with holes where objects in the holdout input have occluded parts of it. You can then merge this over your RGB footage with the Merge node’s mask mode. The final result is your image with holes punched out where occlusion has occurred.
When disabled, DeepHoldout outputs your main RGB image with holes where the objects in the holdout input have occluded parts of it, and an alpha channel for that image.
|