It is well known that counting perfect matchings is tractable in planar graphs (due to Kastelyn).
I am interested in classes of (for lack of a better word) "near" planar graphs (1-planar, bounded genus etc.) where counting perfect matchings is also tractable.