Handling occlusions

[CMCDragonkai]

Was wondering what does imantics do with occlusions?

I've written a custom library that resolves occlusions and when it generates polygon metadata, it actually has a height property represent their z-index layering.

But basically if an object occludes another object. There are 2 ways to go about this.

The background object can substract the foreground object, thus the background polygon is like having a chunk ripped out of it.

The background object includes the foreground object. But only when the background object is completely surrounding the foreground object.

For the purposes of object detection, I did both. The former for partial occlusions. The latter for complete surroundings.