Photo by Milad Fakurian on Unsplash

CoversBR Dataset for Machine Learning

Install DagsHub:

pip install dagshub
Click on copy button to copy content

To stream this data directly on DagsHub

from dagshub.streaming import DagsHubFilesystem

fs = DagsHubFilesystem(".", repo_url="https://dagshub.com/DagsHub-Datasets/covers-br-dataset")

fs.listdir("s3://covers-song-br")
Click on copy button to copy content

Description

CoversBR is the first large audio database with, predominantly, Brazilian music for the tasks of Covers Song Identification (CSI) and Live Song Identifications (LSI). Due to copyright restrictions audios of the songs cannot be made available, however metadata and files of features have public access. Audio streamings captured from radio and TV channels for the live song identification task will be made public. CoversBR is composed of metadata and features extracted from 102298 songs, distributed in 26366 groups of covers/versions, with an average of 3.88 versions per group. The entire collection adds up to a total of approximately 7070 hours and the average song length is 240 seconds (4 minutes).

Additional information

Update frequency

New metadata, songs features files and audio streamings for live song identification will be added as soon as available.

Managed by

Dirceu G Silva

License

The code in this repository is licensed under Apache 2.0

The metadata and the pre-extracted features are licensed under CC BY-NC-SA 4.0

Related datasets

CREMA-D

daps-dataset

UrbanSounds

UrbanSound8K-Labeled Urban Sound Excerpts Dataset

Launch your ML development to new heights with DagsHub

Back to top