GISE-51

GISE-51是由格拉斯哥大学创建的一个包含51个单独声事件类别的数据集，共有16,357个音频片段。该数据集是从FSD50K数据集中提取的，每个音频片段包含一个单独的声事件。数据集的创建过程包括从FSD50K中选择具有单一原始收集标签的音频文件，并通过手动检查确保每个声事件类别的质量。GISE-51数据集适用于声事件识别和检测，旨在解决声事件分类和识别的问题，特别是在声事件的领域特定应用中。

GISE-51 is a dataset created by the University of Glasgow, encompassing 51 distinct sound event classes and a total of 16,357 audio clips. This dataset is extracted from the FSD50K dataset, where each audio clip contains exactly one sound event. The dataset construction process involves selecting audio files with a single original collected label from FSD50K, and verifying the quality of each sound event category through manual inspection. The GISE-51 dataset is tailored for sound event recognition and detection, and aims to address challenges in sound event classification and recognition, particularly in domain-specific sound event applications.