4 files

Making Sense Of Sounds: Data for the machine learning challenge 2018

posted on 29.11.2018, 08:34 by Lara Harris, Oliver Charles Bones
These are the datasets for the Making Sense Of Sounds challenge 2018. The Development download contains the following:

* Folder 'Development' with 1500 audio files, divided equally into five category subfolders: Music, Human, Nature, Effects, and Urban.

* File 'Logsheet_Development.csv, a listing of every file, plus their category, and event type.

The Evaluation download contains the following:

* Folder 'Evaluation' with 500 audio files, not further allocated to any subfolders.

* File 'Logsheet_Evaluation.csv', a listing of all files in the Evaluation dataset.

It should be assumed that all files in this challenge are provided under the licence CC-BY-NC 4.0 (Creative Commons, Attribution Noncommercial1). This is the most restrictive licence of any file in the dataset, though some were also provided under CC02 and CC-BY3. A complete listing of the exact licences and author attributions is contained in 'MSoS_challenge_2018_File_credits_Usage_info_v1-00.zip',

V4 is the latest version of this repository, making the file 'Logsheet_EvaluationMaster.csv' available for download. This matches the format of 'Logsheet_Development.csv' which accompanies the Development dataset. Together these two logsheet files make it possible to check both the main category and the sound event type for every file in the challenge dataset.

See the challenge home page for full details about the dataset and challenge results: