I've used fdupes and refund for exact duplicate finding.

They work shockingly well on large libraries, I've managed to check millions of files consuming several TBs in a matter of hours.