https://github.com/Computational-Cognitive-Musicology-Lab/CoCoPops

CoCoPops is a meta-corpus of melodic and harmonic transcriptions of popular music. CoCoPops has been developed primarily by Nat Condit-Schultz and Claire Arthur in the Computational and Cognitive Musicology Lab, within the Georgia Tech Center for Music Technology. The goal of CoCoPops is to make a large ammount of comparable melodic/harmonic data available in a consistent, standardized format. All CoCoPops files are stored in humdrum format. CoCoPops currently includes of two main sub-corpora, the Billboard subset and the Rolling Stone subset. We plan on continuing to add more as additional datasets and corpora of popular music with melodic transcriptions become available.