Annotations are provided as a .csv
file containing the concert name, audio duration (in seconds), boundary time instants (in seconds) and section labels for every concert in the cross-validation and test datasets.
Refer to Section 2.1 of the paper for more details on the annotation procedure.
train_dataset.csv
- the 20 concert dataset used for cross-validation
test_dataset.csv
- the 2 concert test set