Music Source Restoration with Ensemble Separation and Targeted Reconstruction

TL;DR

This paper introduces a two-stage music source restoration system that combines ensemble separation models with targeted reconstruction, effectively reversing complex production effects to recover original music stems, and achieves top performance on the MSR benchmark.

Contribution

The paper presents a novel two-stage approach integrating ensemble separation and BSRNN-based restoration for music source recovery, outperforming existing methods on the MSR benchmark.

Findings

Outperforms baseline methods on all metrics

Ranks second among all submissions in the MSR challenge

Demonstrates effectiveness of combined separation and restoration approach

Abstract

The Inaugural Music Source Restoration (MSR) Challenge targets the recovery of original, unprocessed stems from fully mixed and mastered music. Unlike conventional music source separation, MSR requires reversing complex production processes such as equalization, compression, reverberation, and other real-world degradations. To address MSR, we propose a two-stage system. First, an ensemble of pre-trained separation models produces preliminary source estimates. Then a set of pre-trained BSRNN-based restoration models performs targeted reconstruction to refine these estimates. On the official MSR benchmark, our system surpasses the baselines on all metrics, ranking second among all submissions. The code is available at https://github.com/xinghour/Music-source-restoration-CUPAudioGroup