tangjjbetsy / atepp Goto Github PK

Hello, is there by chance a colab to infer with your model and checkpoint? Thank you so much

Hello, (I'm only a musician, not an expert in programming) but I got interested in the fact that note durations are more accurate in the midis you provide rather than in Bytedance's original system. I was wondering if you altered their method to achieve those results, if you want to comment on that a little! And I also can't wait to try your system, may I ask, will you release it soon? Thanks for the last response.

Hi, can I get some detail of the copyright in this dataset? According to the papar, it said "we download each track from a corresponding open source audio at YouTube Music".

But I check the csv file that most Youtube music downloaded is provided by the Universal Music Group, and the performed pianists are generally not "died over 70 year" so I assuming their arrangements are under the copyright.

Would you detail the "open source audio" about how this open souce? And same concern for the sheet music. For the pieces in MuseScore, does there scores are non-provided licenses (so have copyright issue), or did provide the opensource licenses? MuseScore is open to public and non-careful management the copyright, so made things become sensitive.

This is a very promising dataset, but unfortunately I need to address the Univerisity's concerns regarding copyright before the further research.

Best regards <3