rpmon/fma-genre-classification

FMA流派分类数据集是Free Music Archive (FMA)的一个子集，包含用于音乐分类任务的音频样本和流派标签。此版本使用FMA的“小”子集，包含8,000个轨道，每个轨道30秒，均匀分布在8种流派中。音频经过预处理，确保格式和采样率一致（16 kHz）。数据集分为训练集（80%）和验证集（20%）。支持的任务包括音频分类和音频特征学习。数据集的结构包括音频文件、流派标签、轨道ID、标题和艺术家名称。创建过程涉及从FMA的small子集中提取数据，并进行预处理。使用数据集时需考虑流派代表性、西方音乐偏见、英语元数据偏见和艺术家代表性等潜在偏见。数据集由rpmon基于原始FMA数据集策划，并遵循Creative Commons许可。

The FMA Genre Classification Dataset is a subset of the Free Music Archive (FMA), containing 8,000 audio samples, each 30 seconds in length, evenly distributed across 8 musical genres. The audio files are in MP3 format and have been resampled to a consistent 16kHz. The dataset is split into training (80%) and validation (20%) sets. The genres included are Electronic, Experimental, Folk, Hip-Hop, Instrumental, International, Pop, and Rock. The dataset is suitable for tasks such as audio classification and audio feature learning. The metadata is in English, and the audio content spans multiple languages. The dataset was curated based on the original FMA dataset and is released under Creative Commons licenses.