imageomics/acoustic-PUUM

这是一个包含未标记、未处理的夏威夷鸟类被动声学录音的数据集，录音来自夏威夷Puu Makaala自然保护区（PUUM）。该数据集旨在用于无监督的音频分析方法、使用现有模型的分类以及其他机器学习和生态学研究目的。此外，数据集还包含天气和鸟类检测的数据框。数据集包括来自9个录音器的被动声学录音（6个沿物候样带放置，3个在koa恢复地点）。录音时间为2025年1月23日至4月3日，每天在黎明合唱期间（6:00-7:00）录制30分钟，以及从7:00到19:00每小时录制15分钟。原始音频文件经过Bird-MixIT模型处理，以从重叠的环境声音中分离出单个鸟类的发声，并使用Perch鸟类声音识别模型（v2.0）进行分类，保留所有概率大于0.01的夏威夷物种检测。

This is a dataset containing unlabelled, unprocessed passive acoustic recordings of Hawaiian birds in the Puu Makaala Natural Area Reserve (PUUM) in Hawaii. It is intended for use in unsupervised audio analysis methods, classification using existing models, and other machine learning and ecology research purposes. Additionally, this dataset contains dataframes with the weather and bird detections. The dataset contains passive acoustic recordings from 9 recorders (6 along a phenology transect, 3 in koa restoration sites). Recordings were taken between January 23 and April 3, 2025, with daily collection of 30 minutes during the dawn chorus (6:00-7:00) and 15 minutes every hour from 7:00 to 19:00. Raw audio files were processed using Bird-MixIT to isolate individual bird vocalizations from overlapping environmental sounds, and classified using the Perch bird sound recognition model (v2.0), retaining all Hawaiian species detections with probability >0.01.