Below, you can find the instructions to download and prepare the datasets used in the benchmark. We suggest to download the datasets in a known directory (e.g., path/to/datasets
) and update the path
parameter in the configs/data_config.json
file with the path to each dataset.
For convenience, we provide a script to download and prepare the datasets. It can be found at data_download/download_data.sh
. The script will create a folder named audio_datasets
and download the datasets in it. Please update the path
parameter in the configs/data_config.json
file with the path to the audio_datasets
folder. If you execute the script to download the datasets, the final command will print the path to the audio_datasets
folder as final output.
wget https://github.com/karoldvl/ESC-50/archive/master.zip
unzip master.zip
rm master.zip
mv ESC-50-master esc50
wget https://zenodo.org/record/1203745/files/UrbanSound8K.tar.gz
tar -xvzf UrbanSound8K.tar.gz
rm UrbanSound8K.tar.gz
mkdir fsd50k
cd fsd50k
wget https://zenodo.org/record/4060432/files/FSD50K.dev_audio.z01
wget https://zenodo.org/record/4060432/files/FSD50K.dev_audio.z02
wget https://zenodo.org/record/4060432/files/FSD50K.dev_audio.z03
wget https://zenodo.org/record/4060432/files/FSD50K.dev_audio.z04
wget https://zenodo.org/record/4060432/files/FSD50K.dev_audio.z05
wget https://zenodo.org/record/4060432/files/FSD50K.dev_audio.zip
wget https://zenodo.org/record/4060432/files/FSD50K.eval_audio.z01
wget https://zenodo.org/record/4060432/files/FSD50K.eval_audio.zip
wget https://zenodo.org/record/4060432/files/FSD50K.ground_truth.zip
wget https://zenodo.org/record/4060432/files/FSD50K.metadata.zip
7z x FSD50K.dev_audio.zip
rm FSD50K.dev_audio.z*
7z x FSD50K.eval_audio.zip
rm FSD50K.eval_audio.z*
unzip FSD50K.ground_truth.zip
unzip FSD50K.metadata.zip
rm FSD50K.ground_truth.zip
rm FSD50K.metadata.zip
mkdir vivae
cd vivae
wget https://zenodo.org/record/4066235/files/VIVAE.zip
unzip VIVAE.zip
rm VIVAE.zip
It needs to have 7z installed given a known error with unzip (error: not enough memory for bomb detection
). If you don't have it, you can install it with:
sudo apt-get install p7zip-full
Then, download and unzip the dataset:
mkdir fma_small
cd fma_small
wget https://os.unil.cloud.switch.ch/fma/fma_small.zip
wget https://os.unil.cloud.switch.ch/fma/fma_metadata.zip
7z x fma_small.zip
7z x fma_metadata.zip
rm fma_small.zip
rm fma_metadata.zip
The split of the dataset is the same as the one used in Musicnn.
mkdir magnatagatune
wget https://mirg.city.ac.uk/datasets/magnatagatune/clip_info_final.csv
wget https://mirg.city.ac.uk/datasets/magnatagatune/mp3.zip.001
wget https://mirg.city.ac.uk/datasets/magnatagatune/mp3.zip.002
wget https://mirg.city.ac.uk/datasets/magnatagatune/mp3.zip.003
wget https://mirg.city.ac.uk/datasets/magnatagatune/annotations_final.csv
cat mp3.zip.* > mp3.zip
unzip mp3.zip
rm mp3.zip
rm mp3.zip.*
wget https://raw.githubusercontent.com/jordipons/musicnn-training/master/data/index/mtt/test_gt_mtt.tsv # download the test split
wget https://raw.githubusercontent.com/jordipons/musicnn-training/master/data/index/mtt/train_gt_mtt.tsv # download the train split
wget https://raw.githubusercontent.com/jordipons/musicnn-training/master/data/index/mtt/val_gt_mtt.tsv # download the validation split
mkdir irmas
cd irmas
wget https://zenodo.org/record/1290750/files/IRMAS-TrainingData.zip
wget https://zenodo.org/record/1290750/files/IRMAS-TestingData-Part1.zip
wget https://zenodo.org/record/1290750/files/IRMAS-TestingData-Part2.zip
wget https://zenodo.org/record/1290750/files/IRMAS-TestingData-Part3.zip
unzip IRMAS-TrainingData.zip
unzip IRMAS-TestingData-Part1.zip
unzip IRMAS-TestingData-Part2.zip
unzip IRMAS-TestingData-Part3.zip
rm IRMAS-TrainingData.zip
rm IRMAS-TestingData-Part1.zip
rm IRMAS-TestingData-Part2.zip
rm IRMAS-TestingData-Part3.zip
mkdir medleydb
cd medleydb
wget https://zenodo.org/record/3464194/files/Medley-solos-DB.tar.gz
wget https://zenodo.org/record/2582103/files/Medley-solos-DB_metadata.csv
mkdir audio
cd audio
mv ../Medley-solos-DB.tar.gz .
tar -xvzf Medley-solos-DB.tar.gz
rm Medley-solos-DB.tar.gz
wget https://zenodo.org/record/1188976/files/Audio_Speech_Actors_01-24.zip
mkdir ravdess
unzip Audio_Speech_Actors_01-24.zip -d ravdess
rm Audio_Speech_Actors_01-24.zip
mv Audio_Speech_Actors_01-24 ravdess
mkdir slurp
cd slurp
wget https://zenodo.org/record/4274930/files/slurp_real.tar.gz
wget https://zenodo.org/record/4274930/files/slurp_synth.tar.gz
tar -xvzf slurp_real.tar.gz
tar -xvzf slurp_synth.tar.gz
rm slurp_real.tar.gz
rm slurp_synth.tar.gz
wget https://raw.githubusercontent.com/pswietojanski/slurp/master/dataset/slurp/devel.jsonl
wget https://raw.githubusercontent.com/pswietojanski/slurp/master/dataset/slurp/test.jsonl
wget https://raw.githubusercontent.com/pswietojanski/slurp/master/dataset/slurp/train.jsonl
mkdir emovo
cd emovo
gdown 1SUtaKeA-LYnKaD3qv87Y5wYgihJiNJAo
unzip emovo.zip
rm emovo.zip