I have a collection of images taken from different windows in buildings pointing out and down. I’m trying to identify the window location on pictures facing the building. So far I’ve tried to just frame this as a regression problem with 3 images as input. See example below.

Example

I want to know if this is an existing problem and if so what is the name for it. I’ve been stuck figuring out what exactly to search. The closest I’ve found is relative camera pose estimation but its not exactly what I want.