ilyassmoummad/Xeno-Canto-6s-16khz
收藏资源简介:
该数据集提供了从Xeno-Canto鸟鸣声数据集中提取的684,744个6秒长的音频片段,涵盖10,127种鸟类。数据集经过预处理,使用CNN14模型选择具有最高鸟鸣声激活的6秒窗口,并将音频下采样至16kHz并转换为Pytorch格式,优化了其在深度学习模型训练中的使用效率。该数据集用于训练特征提取模型,并在研究中评估了模型的少样本学习能力。数据集包含一个CSV文件,记录了每个文件的元数据,如物种代码、科学名称、声音类型、录音细节等。
The Xeno-Canto Bird Sound Dataset is sourced from the Xeno-Canto website and is specifically designed for training deep learning models. This dataset contains 684,744 audio segments, each 6 seconds long, covering 10,127 bird species. The audio data has been preprocessed using the CNN14 model to select 6-second windows with the highest bird sound activation, downsampled to 16kHz, and converted into Pytorch format to optimize training efficiency. The dataset also includes a CSV file with metadata providing detailed information about each bird species. The dataset is compressed into 13 tar files, with a total size of approximately 250GB.




