MAESTRO dataset parser written in C#
Dataset can be acquired from https://magenta.tensorflow.org/datasets/maestro
This program parses information from the MIDI file and translates into following types of data:
label | type | range | source |
---|---|---|---|
Time | Integer | 0 < n | raw time |
TimeDiff | Integer | 0 ≤ n | Time - Previous Time |
Length | Integer | 0 < n | Note Off - Note On |
NoteNum | Integer | 0 ≤ n ≤ 127 | |
Velocity | Integer | 0 ≤ n ≤ 127 |
In the Main() method, you can use 3 methods:
- ReadParseStatistics: print and save the statistics of the dataset(Min, Max, Avg of the labels)
- ReadParseSave: Parse and translate the dataset, and save the labels into results directory
- ReadStandardizeSave: ReadParseSave but standardizes the dataset