The patterns that emerge from a collection of axioms are indeed objective (in that no one else can derive different results from them) and natural (since many natural processes behave according to these patterns regardless of whether anyone has studied them.)
Choosing a set of axioms which gives a good balance of generality and usefulness is an art, so they seem more like inventions. For example, topological spaces are pretty general, so they don't have the nice properties that Hausdorff spaces have, which itself is pretty broad compared to metric spaces, which have a great deal of structure and nice properties. But Euclidean spaces, although having tons of organization, cover a very small portion of useful mathematical objects.
However, since some collections of axioms give a surprising amount of results over other less elegant attempts, one could argue that they would be invented anyway at some point, so they are discovered rather than invented.