Divide and Remaster Utility Tools
Utility tools for the “Divide and Remaster” dataset, introduced as part of the Cocktail Fork problem paper
The DnR dataset is build from three, well-established, audio datasets; Librispeech, Free Music Archive (FMA), and Freesound Dataset 50k (FSD50K). We offer our dataset in both 16kHz and 44.1kHz sampling-rate along time-stamped annotations for each of the classes (genre for ‘music’, audio-tags for ‘sound-effects’, and transcription for ‘speech’). We provide below more informations on how the dataset is build and what it’s consists of exactly. We also go over the process of building the dataset from scratch for the cases it needs to.
- Dataset Overview
- Get the DnR Dataset
- Resources and Support
The Divide and Remaster (DnR) dataset is a dataset aiming at providing research support for a relatively unexplored case of source separation with mixtures involving music, speech, and sound-effects (SFX) as their sources. The dataset is build from three, well-established, datasets. Consequently if one wants to build DnR from scratch, the aforementioned datasets will have to be downloaded first. Alternatively, DnR is also available on Zenodo
Get the DnR Dataset
In order to obtain DnR, several options are available depending on the task at hand:
- DnR-HQ (44.1kHz) is available on Zenodo at the following or simply run:
link to the Zenodo dataset coming soon ...
- Alternatively, if DnR-16kHz is needed, please first download DnR-HQ locally. You can then downsample the dataset (either in-place or not) by cloning the dnr-utils repository and running:
python dnr_utils.py --task=downsample --inplace=True
Building DnR From Scratch
In the section, we go over the DnR building process. Since DnR is directly drawn from *FSD50K*, *LibriSpeech*/*LibriVox*, and *FMA, we first need to download these datasets. Please head to the following links for more details on how to get them:
Please note that for FMA, the medium set only is required. In addition to the audio files, the metadata should also be downloaded. For LibriSpeech DnR uses dev-clean, test-clean, and train-clean-100. DnR will use the folder structure as well as metadata from LibriSpeech, but ultimately will build the LibriSpeech-HQ dataset off the original LibriVox mp3s, which is why we need them both for building DnR.
After download, all four datasets are expected to be found in the same root directory. Our root tree may look something like that. As the standardization script will look for specific file name, please make sure that all directory names conform to the ones described below:
root ├── fma-medium │ ├── fma_metadata │ │ ├── genres.csv │ │ └── tracks.csv │ ├── 008 │ ├── 008 │ ├── 009 │ └── 010 │ └── ... ├── fsd50k │ ├── FSD50K.dev_audio │ ├── FSD50K.eval_audio │ └── FSD50K.ground_truth │ │ ├── dev.csv │ │ ├── eval.csv │ │ └── vocabulary.csv ├── librispeech │ ├── dev-clean │ ├── test-clean │ └── train-clean-100 └── librivox ├── 14 ├── 16 └── 17 └── ...
Once all four datasets are downloaded, some standardization work needs to be taken care of. The standardization process can be be executed by running
standardization.py, which can be found in the dnr-utils repository. Prior to running the script you may want to install all the necessary dependencies included as part of the
pip install -r requirements.txt. Note:
ffmpeg under its hood, a system install of
fmmpeg is thus required. Please see pydub’s install instructions for more information. The standardization command may look something like:
python standardization.py --fsd50k-path=./FSD50K --fma-path=./FMA --librivox-path=./LibriVox --librispeech-path=./LibiSpeech --dest-dir=./dest --validate-audio=True
DnR Dataset Compilation
Once the three resulting datasets are standardized, we are ready to finally compile DnR. At this point you should already have cloned the dnr-utils repository, which contains two key files:
config.pycontains some configuration entries needed by the main script builder. You want to set all the appropriate paths pointing to your local datasets and ground truth files in there.
- The compilation for a given set (here,
eval) can be executed with
compile_dataset.py, for example by running the following commands for each set:
python compile_dataset.py with cfg.train
python compile_dataset.py with cfg.val
python compile_dataset.py with cfg.eval
Some known bugs and issues that we’re aware. if not listed below, feel free to open a new issue here:
If building from scratch,
pydubwill fail at reading 15 mp3 files from the FMA medium-set and will return the following error:
mp3 @ 0x559b8b084880] Failed to read frame size: Could not seek to 1026.
If building DnR from scratch, the script may return the following error, coming from
Audio must be have length greater than the block size. That’s because some audio segment, especially SFX events, may be shorter than 0.2 seconds, which is the minimum sample length (window) required by
pyloudnormfor normalizing the audio. We just ignore these segments.
Contact and Support
Have an issue, concern, or question about DnR or its utility tools ? If so, please open an issue
For any other inquiries, feel free to shoot an email at: [email protected], my name is Darius Petermann 😉