mdeff / fma

FMA: A Dataset For Music Analysis

Home Page:https://arxiv.org/abs/1612.01840

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Understanding Column Labels (01, 02, 03, etc.) in FMA Dataset

darren7753 opened this issue · comments

Hello, I'm currently working on my thesis about music genre classification using the FMA dataset. However, I'm having trouble understanding the meaning of these columns labeled 01, 02, 03, and so on. Could you please assist me with this? I would greatly appreciate it.

messageImage_1697303257294

In case an answer to this would still be helpful: MFCC produces a sequence of coefficients, numbered 01, 02, ..., for each small time window of a track. To generate that table, the programmers calculated the first 20 of these coefficients, and then took their mean, variance, etc. over the whole track. So the column ("mfcc", "median", 02) gives the median of the 2nd MFC coefficient over the whole track.

The situation's the same for the other features, and many of those are easier to understand than MFCC. For example, there are 12 chroma_cqt features for each time window, representing the strength of the 12 chromatic pitch classes in that window. The summary statistics for those features tell you about the strength of the 12 pitch classes over the whole track.