I'm looking to develop an assistant software that'll perform as an overlay on top of a game window that will detect certain images and outline them. I need the program to detect exact 2D images/icons (e.g. currency items in the inventory). Since the 2d icons I want to detect on screen don't vary at all (due to being UI elements), and always appear exactly the same (aside from pixel resolution possibly being different based on the game window's resolution), I'm wondering if training a model on custom datasets would be overkill.
I'm asking for advice on what approach i should take, and whether computer vision is the way to go for my purpose.