Tensorflow2 implementation of various music auto-tagging models

Evaluation of CNN-based Automatic Music Tagging Models

Minz Won, Andres Ferraro, Dmitry Bogdanov, and Xavier Serra

SMC, 2020

Reference

• arxiv

• pytorch

Available Models

• Sample-level CNN : Sample-level Deep Convolutional Neural Networks for Music Auto-tagging using Raw Waveforms, Lee et al., 2017 pdf
• Sample-level CNN + Squeeze-and-excitation : Sample-level CNN Architectures for Music Auto-tagging Using Raw Waveforms, Kim et al., 2018 arxiv, code
• Music-SincNet : Music Auto-tagging with Learning Filter Banks, Lee and Shin, 2020 pdf
• Harmonic-CNN : Data-driven Harmonic filters for Audio Representation Learning, Won et al., 2020 pdf, code

Upcoming Models

• Self-attention

Usage

• Requirements

conda env create -n {ENV_NAME} --file environment.yaml
conda activate {ENV_NAME}

• Preprocessing

python -u preprocess.py run ../dataset
python -u split.py run ../dataset

• Training

python main.py

• Options

'--gpu', type=str, default='0'
'--encoder_type', type=str, default='HC', choices=['HC', 'MS', 'SC']
'--block', type=str, default='rese', choices=['basic', 'se', 'res', 'rese']