crokobit / indievox-dataset Goto Github PK

iNDIEVOX open datasets, you can play machine learning algorithms with these datasets.

You can check out and listen to the music by song id, using the url format like: https://www.indievox.com/song/song_id. for example: https://www.indievox.com/song/1

• Valence-Arousal Dataset Song ID List - Top 200 song ids are for testing and the other 800 song ids are for training.
• Valence Training Dataset - Top 68 columns are audio features(use pyaudio) and the last column is valence value.
• Valence Testing Dataset - Top 68 columns are audio features(use pyaudio) and the last column is valence value.
• Arousal Training Dataset - Top 68 columns are audio features(use pyaudio) and the last column is arousal value.
• Arousal Testing Dataset - Top 68 columns are audio features(use pyaudio) and the last column is arousal value.

• Relax Song ID List - Top 10 song ids are for testing and the other 60 song ids are for training.
• Happy Song ID List - Top 10 song ids are for testing and the other 60 song ids are for training.
• Excited Song ID List - Top 10 song ids are for testing and the other 60 song ids are for training.
• Blue Song ID List - Top 10 song ids are for testing and the other 60 song ids are for training.
• Sad Song ID List - Top 10 song ids are for testing and the other 60 song ids are for training.
• Angry Song ID List - Top 10 song ids are for testing and the other 60 song ids are for training.
• Emotion Training Dataset - Top 68 columns are audio features(use pyaudio) and the last column is emotion category number.
• Emotion Testing Dataset - Top 68 columns are audio features(use pyaudio) and the last column is emotion category number.

• Buy Together Discs - Each row is transaction that discs buys together, each column is the disc id number.
• Buy Together Songs - Each row is transaction that songs buys together, each column is the song id number.

We play algorithms like Gaussian SVM and Ridge Regression to train ML models from the datasets. The algorithm library we use is Fuku-ML, and feature extraction library is PyAudio.