SHS-YT
收藏资源简介:
SHS-YT数据集由柏林洪堡大学的研究团队创建,旨在评估现有版本识别模型在处理YouTube上的翻唱歌曲时的鲁棒性。该数据集包含900个版本,数据来源于YouTube平台,涵盖了西方流行音乐的广泛样本。数据集的创建过程包括从YouTube上检索候选版本、使用多模态不确定性采样方法筛选版本,并通过众包平台Mechanical Turk进行人工标注。该数据集的应用领域主要集中在音乐检索和版权侵权检测,旨在解决现有模型在处理YouTube上的翻唱歌曲时可能遇到的挑战。
The SHS-YT dataset was created by a research team from Humboldt-Universität zu Berlin, aiming to evaluate the robustness of existing version identification models when handling cover songs on YouTube. This dataset includes 900 versions sourced from the YouTube platform, covering a broad range of Western popular music samples. The dataset's creation process involves retrieving candidate versions from YouTube, screening these versions using the multimodal uncertainty sampling method, and conducting manual annotation via Amazon Mechanical Turk. The primary application domains of this dataset are music retrieval and copyright infringement detection, which aims to address the challenges that existing models may encounter when processing cover songs on YouTube.

- 1On the Robustness of Cover Version Identification Models: A Study Using Cover Versions from YouTube柏林洪堡大学图书馆与信息科学学院 · 2025年



