Published May 21, 2021 | Version v1
Dataset Open

MATS - Multi-Annotator Tagged Soundscapes

  • 1. Tampere University

Description

This is a dataset containing audio tags for a number of 3930 audio files of the TAU Urban Acoustic Scenes 2019 development dataset (airport, public square, and park). The files were annotated using a web-based tool, with multiple annotators providing labels for each file. 

 

The dataset contains annotations for 3930 files, annotated with the following tags:

  • announcement jingle
  • announcement speech
  • adults talking
  • birds singing
  • children voices
  • dog barking
  • footsteps
  • music
  • siren
  • traffic noise

The annotation procedure and processing is presented in the paper:

Irene Martin-Morato, Annamaria Mesaros. What is the ground truth? Reliability of multi-annotator data for audio tagging, 29th European Signal Processing Conference, EUSIPCO 2021 

 

The dataset contains the following:

  • raw annotations provided by 133 annotators, multiple opinions per audio file

        MATS_labels_full_annotations.yaml 

        content formatted as: 

               - filename: file1.wav 
                 annotations:
                 - annotator_id: ann_1
                   tags:
                   - tag1
                   - tag2
                 - annotator_id: ann_3
                   tags:
                   - tag1
               - filename: file3.wav 
                ...

            

  • processed annotations using different methods, as presented in the accompanying paper

           MATS_labels_majority_vote.csv
           MATS_labels_union.csv
           MATS_labels_mace100.csv   
           MATS_labels_mace100_competence60

           content formatted as: 

           filename [tab] tag1,tag2,tag3


The audio files can be downloaded from https://zenodo.org/record/2589280 and are covered by their own license.

 

 

Files

LICENSE.txt

Files (2.5 MB)

Name Size Download all
md5:03fc4e8585b91d54740a6f6f092bc708
1.5 kB Preview Download
md5:d1cef455e5ea143b9c0ee5990858e3ba
1.5 MB Download
md5:d7eff5db9e2cdac8305b91cf0ac4df21
255.1 kB Preview Download
md5:11bc936a80f2474a54e5968a7970510f
255.0 kB Preview Download
md5:aac9cf2b7d49d607e98a1573825f29c9
203.3 kB Preview Download
md5:c48907185b35cce6b226d1ad564a7b36
290.5 kB Preview Download

Additional details

Funding

Teaching machines to listen 332063
Research Council of Finland

References

  • Irene Martin-Morato, Annamaria Mesaros. What is the ground truth? Reliability of multi-annotator data for audio tagging, 29th European Signal Processing Conference, EUSIPCO 2021, https://arxiv.org/abs/2104.04214