This is a deep-learning-based tool to extract instrumental track from mixture audio.
pip install -r requirements.txt
Download the latest version from here.
The following code splits the mixture audio into an instrumental track and a vocal track. These tracks are saved as *_Instrumental.wav and *_Vocal.wav.
python inference.py --input path/to/mixture/audio
python inference.py --input path/to/mixture/audio --gpu 0
sudo apt install soundstretch
dataset/
+- instrumentals/
| +- 01_foo_inst.wav
| +- 02_bar_inst.mp3
| +- ...
+- mixtures/
+- 01_foo_mix.wav
+- 02_bar_mix.mp3
+- ...
python augment.py -i dataset/instrumentals -m dataset/mixtures -p -1
python augment.py -i dataset/instrumentals -m dataset/mixtures -p 1
python train.py -i dataset/instrumentals -m dataset/mixtures -M 0.5 -g 0
- [1] Jansson et al., "Singing Voice Separation with Deep U-Net Convolutional Networks", https://ismir2017.smcnus.org/wp-content/uploads/2017/10/171_Paper.pdf
- [2] Takahashi et al., "Multi-scale Multi-band DenseNets for Audio Source Separation", https://arxiv.org/pdf/1706.09588.pdf
- [3] Liutkus et al., "The 2016 Signal Separation Evaluation Campaign", Latent Variable Analysis and Signal Separation - 12th International Conference