LightSAFT: Lightweight Latent Source Aware Frequency Transform for Source Separation

TL;DR

This paper introduces LightSAFT-Net, a lightweight and improved conditioned source separation model that achieves competitive performance with fewer parameters, enhancing the original LaSAFT-Net architecture.

Contribution

The paper proposes LightSAFT-Net, a more efficient version of LaSAFT-Net, and further improves it by integrating TFC-TDF blocks, resulting in better performance with fewer parameters.

Findings

LightSAFT-Net performs well in the Music Demixing Challenge.

Enhanced LightSAFT-Net outperforms the original with fewer parameters.

The model achieves competitive SDR performance.

Abstract

Conditioned source separations have attracted significant attention because of their flexibility, applicability and extensionality. Their performance was usually inferior to the existing approaches, such as the single source separation model. However, a recently proposed method called LaSAFT-Net has shown that conditioned models can show comparable performance against existing single-source separation models. This paper presents LightSAFT-Net, a lightweight version of LaSAFT-Net. As a baseline, it provided a sufficient SDR performance for comparison during the Music Demixing Challenge at ISMIR 2021. This paper also enhances the existing LightSAFT-Net by replacing the LightSAFT blocks in the encoder with TFC-TDF blocks. Our enhanced LightSAFT-Net outperforms the previous one with fewer parameters.Conditioned source separations have attracted significant attention because of their flexibility, applicability and extensionality. Their performance was usually inferior to the existing approaches, such as the single source separation model. However, a recently proposed method called LaSAFT-Net has shown that conditioned models can show comparable performance against existing single-source separation models. This paper presents LightSAFT-Net, a lightweight version of LaSAFT-Net. As a baseline, it provided a sufficient SDR performance for comparison during the Music Demixing Challenge at ISMIR 2021.