Coursework 1 : Music Informatics (2021) Queen Mary University of London
Dynamic Programming implementation and evaluation of Dan Ellis (2007) Beat Tracking algorithm [1].
- python3
- librosa
- mir_eval
To run the program against a single file:
$ make track-beat AUDIO_FILE=<path-to-audio-file>
The program is also callable from a python environment as follows:
beats = beatTracker(<path-to-audio-file>)
To run the program against the entire Ballroom dataset:
make track-all
To evaluate a single run you must have your estimates and your ground truth sequences in local files and the evaluator can be evoked by running:
$ make eval-one BEAT_ESTIMATE=<path-to-estimate-sequence> BEAT_GROUND_TRUTH=<path-to-ground-truth>
To evaluate all outputs in the given dataset and produce relevant box plots:
$ make eval-all