Download

The dataset is composed of two subsets:

Links to the zenodo repos: DESED_synthetic, DESED_real

 

After downloading the data (see below) you should have this tree:

├── dcase2019
│   ├── dataset
│   │   ├── audio
│   │   │   ├── eval
│   │   │   │   ├── 500ms
│   │   │   │   ├── 5500ms
│   │   │   │   ├── 9500ms
│   │   │   │   ├── distorted_clipping
│   │   │   │   ├── distorted_drc
│   │   │   │   ├── distorted_highpass_filter
│   │   │   │   ├── distorted_lowpass_filter
│   │   │   │   ├── distorted_smartphone_playback
│   │   │   │   ├── distorted_smartphone_recording
│   │   │   │   ├── fbsnr_0dB
│   │   │   │   ├── fbsnr_15dB
│   │   │   │   ├── fbsnr_24dB
│   │   │   │   ├── fbsnr_30dB
│   │   │   │   ├── ls_0dB
│   │   │   │   ├── ls_15dB
│   │   │   │   └── ls_30dB
│   │   │   ├── train
│   │   │   │   ├── synthetic
│   │   │   │   ├── unlabel_in_domain
│   │   │   │   └── weak
│   │   │   └── validation
│   │   └── metadata
│   │       ├── eval
│   │       ├── train
│   │       └── validation
│   └── src
├── real_data                                   (subpart of dcase2019)
│   ├── audio
│   │   ├── train
│   │   │   ├── unlabel_in_domain
│   │   │   └── weak
│   │   └── validation
│   ├── metadata
│   │   ├── train
│   │   └── validation
│   └── src
└── synthetic
    ├── audio
    │   ├── eval
    │   │   ├── distorted_fbsnr_30dB            (6 subfolders for each distortion, audio are directly given because a matlab code has been used to generate them) 
    │   │   └── soundbank                       (Raw (bank of) data that can be used to create synthetic data)
    │   │       ├── background                  (2 subfolders, youtube and freesound)
    │   │       ├── background_long             (5 subfolders)
    │   │       ├── foreground                  (18 subfolders)
    │   │       ├── foreground_on_off           (10 subfolders)
    │   │       └── foreground_short            (5 subfolders)
    │   └── train
    │       ├── soundbank                       (Raw (bank of) data that can be used to create synthetic data)
    │       │   ├── background
    │       │   │   └── sins                    (Has to be downloaded by: get_background_training.py)
    │       │   └── foreground                  (14 subfolders)
    │       └── synthetic
    ├── metadata
    │   ├── eval
    │   │   └── soundscapes                     (metadata to reproduce the wav files used in dcase2019)
    │   │       ├── 500ms
    │   │       ├── 5500ms
    │   │       ├── 9500ms
    │   │       ├── fbsnr_0dB
    │   │       ├── fbsnr_15dB
    │   │       ├── fbsnr_24dB
    │   │       ├── fbsnr_30dB
    │   │       ├── ls_0dB
    │   │       ├── ls_15dB
    │   │       └── ls_30dB
    │   └── train
    │       └── soundscapes                     (metadata to reproduce the wav files used in dcase2019)
    └── src                                     (Source code to regenerate the dcase2019 dataset or generate new mixtures)

Comments are closed.

  • Share this

    Facebooktwitterlinkedinmail