FSD-2k

Created By Félix Gontier and Mathieu Lagrange, LS2N, CNRS, Ecole Centrale Nantes Contact : [e-mail] If used for research, please refer to: <pre>@article{gontier2021training, title={Polyphonic training set synthesis improves self-supervised urban sound classification}, author={Félix Gontier and Vincent Lostanlen, and Mathieu Lagrange and Nicolas Fortin and Jean-Francois Petiot and Catherine Lavandier}, journal={The Journal of the Acoustical Society of America}, year={2021}, publisher={Acoustical Society of America} } </pre> FSD-2k contains about 200 monophonic audio clips collected from online resources, which are unrelated to the city of Lorient: Freesound for birds and traffic and Librispeech for voice.<br> <br> The total duration of the dataset is of the order of 2.4k seconds, i.e., 40 minutes. Each audio samples are cut into one or several 3 seconds parts, each resulting into spectrograms of size 23x29, leading to a dataset of 609 spectrograms. Low volume amorphic background noise recordings is added and the cut audio sample is centered within the 3 seconds if shorter. &gt;&gt; import numpy as np &gt;&gt; s=np.load('FSD-2k_train_spectralData.npy') &gt;&gt; print(s.shape) (609, 23, 29) The three dimensions respectively corresponds to the sceneId, the frameId (time), and the spectralId (frequency). &gt;&gt; a=np.load('FSD-2k_train_presence.npy') &gt;&gt; print(a.shape) (609, 16, 3) The 3 dimensions corresponds to the sceneId, the frameId (time), the sourceId (traffic, voice, birds). Annotation is provided as a binary indicator of source presence for one second, that is 8 consecutive 125 ms frames with a hop of one frame.本数据集由Félix Gontier与Mathieu Lagrange创建，依托LS2N实验室、法国国家科学研究中心（CNRS）及南特中央理工学院（Ecole Centrale Nantes）完成。联系方式：[e-mail]。若用于科研工作，请引用以下文献： @article{gontier2021training, title={多声部训练集合成优化自监督城市声分类}, author={Félix Gontier and Vincent Lostanlen, and Mathieu Lagrange and Nicolas Fortin and Jean-Francois Petiot and Catherine Lavandier}, journal={The Journal of the Acoustical Society of America}, year={2021}, publisher={Acoustical Society of America} } FSD-2k 数据集包含约200条单声道音频片段，均采集自公开资源，且与洛里昂市无关：其中鸟类与交通声样本来自Freesound平台，人声样本来自Librispeech平台。 该数据集总时长约2400秒，即40分钟。每条音频样本被裁剪为一段或多段3秒的片段，每段片段均生成尺寸为23×29的语谱图（spectrogram），最终总计得到609张语谱图。处理过程中会添加低音量的无定形背景噪声；若原音频短于3秒，则将裁剪后的音频片段居中放置在3秒的时长区间内。 >>> import numpy as np >>> s=np.load('FSD-2k_train_spectralData.npy') >>> print(s.shape) (609, 23, 29) 该数组的三个维度依次对应场景ID（sceneId）、帧ID（时间维度，frameId）与频谱ID（频率维度，spectralId）。 >>> a=np.load('FSD-2k_train_presence.npy') >>> print(a.shape) (609, 16, 3) 该数组的三个维度依次对应场景ID、帧ID（时间维度）与声源ID（涵盖交通声、人声、鸟类声三类）。标注以二进制指示器形式给出，用于表示1秒时长内的声源存在状态，该1秒区间包含8个连续的125毫秒帧，且帧步长为1帧。