| Author | Iris Yuping Ren |
| Affiliation | Utrecht University |
| Code | github |
I made a video to visualise the features of a song I like. Features: mfcc, chroma, and beat.
Feature extraction was done using the librosa package in python. Then the feature vector/matrices were exported as pickle files.
A separate script was written to load the extracted features. Then the features were mapped to a scatter plot animation:
mfcc[0-2] → colour[r, g, b]
chroma → different y values
beat → marker size
The song is a 6-mins long piano improvisation of a friend in the .m4a format.
Python isn't the prettiest and fastest for visualisation but it can show the essential information (and lots of MIR algorithms are in python!)
Running out of battery towards the end of the hack. More automatisation needs to be done…