I’ve been slowly working towards making a Pokemon Snap! style game, and the first step for that is being able to check if a creature is actually even in the photo.

This first prototype uses rendertextures to determine how in-frame the character is, and how much is occluded by other objects such as grass, trees etc.
The reason for this is that it’s much more precise than using something like, say, raycasts. I’ll also be able to extend this method quite a bit by adding more data to the passes.

This will allow me to do things like, for example, detecting which specific parts of the creature are visible.

I can also use this method to calculate for multiple characters in a scene by colour coding them. Doing this, I can get away with using a single occlusion pass, cutting down on a lot of excess passes.

Currently this is all happening in real-time. The next step is to get it functioning as a “photo” which captures the image and the relevant data for each creature in frame when the player snaps the shot.