Sound Demixing Challenge 2023 Music Demixing Track Technical Report: TFC-TDF-UNet v3

Minseok Kim, Jun Hyung Lee, Soonyoung Jung

6/15/2023

cs.SDcs.LGcs.MM

No official code found. Best implementation: kuielab/sdx23 (96 stars, MIT, PyTorch, last commit 2023-07-14).

Abstract

In this report, we present our award-winning solutions for the Music Demixing Track of Sound Demixing Challenge 2023. First, we propose TFC-TDF-UNet v3, a time-efficient music source separation model that achieves state-of-the-art results on the MUSDB benchmark. We then give full details regarding our solutions for each Leaderboard, including a loss masking approach for noise-robust training. Code for reproducing model training and final submissions is available at github.com/kuielab/sdx23.

View on arXiv View PDF

Code Implementations(1)

kuielab/sdx23Community100%

PyTorchMIT

Sound Demixing Challenge 2023

9611PythonApr 19, 20233 years agoMIT

music-source-separationpytorchsource-separation

Similar Papers

Benchmarks and leaderboards for sound demixing tasks

88% match1 repo8562023

Mamba2 Meets Silence: Robust Vocal Source Separation for Sparse Regions

76% match2025

The ICASSP SP Cadenza Challenge: Music Demixing/Remixing for Hearing Aids

76% match1 repo1942023

Summary of The Inaugural Music Source Restoration Challenge

76% match2026

Pac-HuBERT: Self-Supervised Music Source Separation via Primitive Auditory Clustering and Hidden-Unit BERT

76% match2023

SAM Audio: Segment Anything in Audio

75% match2025

Cite this paper

@article{kim2023sound,
  title  = {Sound Demixing Challenge 2023 Music Demixing Track Technical Report: TFC-TDF-UNet v3},
  author = {Minseok Kim and Jun Hyung Lee and Soonyoung Jung},
  year   = {2023},
  eprint = {2306.09382},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2306.09382v3}
}

Add a badge to your README

Shows whether official code is linked for this paper.

[![Code of Paper](https://codeofpaper.com/badge/2306.09382.svg)](https://codeofpaper.com/paper/2306.09382)