A Convolutional Neural Network for drum signal separation from full mixes.
For a quick demontration, do any of the following:
- run the network on your own recording by:
1. cloning this repository on your machine
2. in a python3.7 environment, installing the required packagespip install -r requirements.txt
3. running the program on your own WAVE or mp3 filepython3 evaluate.py <input_audio_path> <output_audio_path>
; or using the provided file snippet.wav (which was not used during training) - watch this short video https://drive.google.com/file/d/1XmyjkJd7u3MMCC3NKeiOHhJ2nZe7ohR_/view?usp=drive_link
- open up the evaluate.ipynb notebook with an IPython kernel and play the audio in the cell outputs
Method based on this this paper by K. W. E. Lin, B. Balamurali, E. Koh, S. Lui, and D. Herremans: https://arxiv.org/abs/1812.01278
... and this tutorial by Ale Koretzky: https://towardsdatascience.com/audio-ai-isolating-vocals-from-stereo-music-using-convolutional-neural-networks-210532383785
Network trained on the MUSDB18-HQ dataset: https://sigsep.github.io/datasets/musdb.html#musdb18-hq-uncompressed-wav