Occlusion culling clarification

Does occlusion culling work on an object or poly level? I’m working on a scene where I could combine meshes to save on draw calls, but I’d still like to use occlusion culling for more savings. However, the meshes are placed in such a manor that there is no view where at least one object won’t be seen.

Graphic example: Top down room with green marks representing objects.

If I were to combine all the green objects into a single mesh, I could save significant draw calls. However, there are objects in every room, meaning object-based occlusion would be of no use once all meshes are combined into a single object.

occlussion is per object

so you better don’t combine them once they are more than a few occlussion cells in size