Wave-U-Net: A Multi-Scale Neural Network for End-to-End Audio Source Separation

8 Jun 2018Daniel StollerSebastian EwertSimon Dixon

Models for audio source separation usually operate on the magnitude spectrum, which ignores phase information and makes separation performance dependant on hyper-parameters for the spectral front-end. Therefore, we investigate end-to-end source separation in the time-domain, which allows modelling phase information and avoids fixed spectral transformations... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Music Source Separation MUSDB18 STL2 SDR (vocals) 3.253060 # 3
Music Source Separation MUSDB18 STL2 SDR (drums) 4.223755 # 3
Music Source Separation MUSDB18 STL2 SDR (other) 2.251890 # 2
Music Source Separation MUSDB18 STL2 SDR (bass) 3.208830 # 2