A Semantically Consistent Dataset for Data-Efficient Query-Based Universal Sound Separation

TL;DR

This paper introduces Hive, a high-quality synthetic dataset for query-based sound separation, demonstrating that training on pure, semantically consistent data improves model performance and generalization with less data.

Contribution

The authors developed an automated pipeline to create a high-purity dataset, Hive, enabling more data-efficient training of sound separation models with improved robustness.

Findings

Models trained on Hive achieve competitive accuracy with much less data.

Hive-trained models show strong zero-shot generalization.

Prioritizing data purity enhances robustness and reduces computational costs.

Abstract

Query-based universal sound separation is fundamental to intelligent auditory systems, aiming to isolate specific sources from mixtures. Despite recent advances, existing methods continue to suffer from residual interference in complex acoustic scenes. This performance limitation stems largely from a data bottleneck: in-the-wild datasets contain weak labels and severe co-occurrence of events. These flaws induce models to learn spurious correlations between background noise and target categories instead of robust acoustic features. To address this, we propose an automated pipeline that eliminates co-occurrence of events by mining high-purity single-event segments from in-the-wild datasets via a semantically consistent synthesis protocol. Utilizing this pipeline, we constructed Hive, a high-quality synthetic dataset comprising 2.4k hours of raw audio. Experimental results demonstrate that, compared with the state-of-the-art model SAM-Audio which was trained on a huge dataset $\sim$500 times larger than Hive, certain open-source models trained on Hive achieve competitive separation accuracy and perceptual quality. Moreover, these models exhibited remarkable zero-shot generalization on out-of-distribution evaluation benchmarks. These findings highlight that prioritizing purity of supervised signals enables significant data efficiency, offering a new paradigm for training robust auditory foundation models with reduced computational costs. Code and dataset are available at https://shandaai.github.io/Hive.