The current scripts searches on iTunes API for artists provided in a list (see step 2 under instructions) and downloads a 30 second preview of each song from those artists.
The resulting dataset could be used for training Deep Learning models in singer identification tasks.
- Clone the current repository:
git clone https://github.com/aitor-alvarez/MIR-song-dataset-collection.git - Create a file with the artists/performers list. The file should be named
artists.txtand contain a single column with the first row (as header) named artist. Place all your artists' names below this header. - Execute the following command to parse the list of artists and search iTunes API for their songs:
python main.py -a artists.txt. This will result in the creation of a file named dataset.csv that will contain all songs from those artists. - Feel free to edit dataset.csv if you want to exclude songs.
- To download the 30 second previews execute:
python main.py -d - All previews will be downloaded into a subfolder within this repository named
songs/