ryanleeme17/free-music-archive-retrieval

Free-Music-Archive-Retrieval（FMAR）是一个为了提高版权侵权检测能力而设计的结构化数据集。它旨在测试模型基于5秒音频片段（查询）识别歌曲的能力。该数据集包含了经过背景噪声添加、音高转换和均衡器平衡等修改的对抗性查询。数据来源于Hugging Face上的benjamin-paine/free-music-archive-small集合，共有7916条音频轨道，平均每条轨道约30秒，覆盖多种音乐风格，以确保音乐风格的多样性。Free-Music-Archive-Retrieval (FMAR) is a structured dataset designed to enhance copyright infringement detection by testing a models ability to identify songs based on 5-second audio clips, or queries. The dataset includes adversarial queries that have been modified with added background noise, pitch shifting, and EQ balancing. The data is sourced from the benjamin-paine/free-music-archive-small collection on Hugging Face, comprising 7,916 audio tracks, each with an average duration of approximately 30 seconds, covering multiple genres to ensure a diverse representation of musical styles.