D3Net: Densely connected multidilated DenseNet for music source separation

TL;DR

This paper introduces D3Net, a novel CNN architecture with multi-dilated convolutions and dense connections, effectively modeling long-term dependencies in music source separation and achieving state-of-the-art results.

Contribution

D3Net combines multi-dilated convolutions with DenseNet to model multi-resolution data efficiently, avoiding aliasing and improving separation performance.

Findings

Achieves an SDR of 6.01 dB on MUSDB18

Outperforms previous CNN-based methods in source separation

Effectively models long-term dependencies in audio signals

Abstract

Music source separation involves a large input field to model a long-term dependence of an audio signal. Previous convolutional neural network (CNN)-based approaches address the large input field modeling using sequentially down- and up-sampling feature maps or dilated convolution. In this paper, we claim the importance of a rapid growth of a receptive field and a simultaneous modeling of multi-resolution data in a single convolution layer, and propose a novel CNN architecture called densely connected dilated DenseNet (D3Net). D3Net involves a novel multi-dilated convolution that has different dilation factors in a single layer to model different resolutions simultaneously. By combining the multi-dilated convolution with DenseNet architecture, D3Net avoids the aliasing problem that exists when we naively incorporate the dilated convolution in DenseNet. Experimental results on MUSDB18 dataset show that D3Net achieves state-of-the-art performance with an average signal to distortion ratio (SDR) of 6.01 dB.