Measure how diverse a dataset of point clouds is

Question

I have a dataset of N point clouds of the same size, i.e objects represented by M 3d points:

Object 1: [(x1, y1, z1), ..., (xM, yM, zM)]
.
.
.
Object N: [(x1, y1, z1), ..., (xM, yM, zM)]

There are some duplications in the dataset. I'm sampling sub-sets of from the dataset, and want to tell which sub-set has the least duplications.

How can i measure the how diverse each sub-set is?

Thanks!