MeerKAT

MeerKAT是由马克斯·普朗克动物行为研究所等机构合作创建的大型生物声学数据集，包含来自自由活动的猫鼬的音频记录，总时长超过1068小时，其中184小时具有时间解析的12种声音类型标签。该数据集通过生物记录器收集，具有毫秒级的时间分辨率，是目前公开的最大的非人类陆地哺乳动物声学数据集。创建过程中，研究人员通过精心的录音和标注确保了数据的质量和准确性。MeerKAT数据集主要用于生物声学模型的预训练和微调，特别是在处理稀疏和不平衡数据方面，为生物声学研究提供了一个新的参考标准，有助于解决动物行为、生态和保护等领域的关键问题。

MeerKAT is a large-scale bioacoustic dataset co-developed by institutions including the Max Planck Institute of Animal Behavior and other collaborators. It contains audio recordings collected from free-ranging meerkats, with a total duration exceeding 1,068 hours, among which 184 hours are annotated with time-resolved labels for 12 distinct sound categories. Collected via bio-loggers, the dataset features millisecond-level temporal resolution, making it the largest publicly available acoustic dataset for non-human terrestrial mammals to date. During its creation, researchers ensured data quality and accuracy through rigorous recording and annotation workflows. The MeerKAT dataset is primarily designed for pre-training and fine-tuning bioacoustic models, particularly for handling sparse and imbalanced data, serving as a new benchmark for bioacoustic research and aiding in addressing critical questions in fields such as animal behavior, ecology, and conservation.