I prefer the read and compare method used by https://github.com/ssokolow/fastdupes as it does not have to read everything to find the exact duplicates which the hashing method must.