BirdCLEF 2018

BirdClef 2018 是一个基于 Xeno-canto 网络贡献的鸟类音景数据集。训练集包含 36,496 个录音，覆盖中美洲和南美洲的 1500 种物种（文献中最大的生物声学数据集）。总共有大约 68 小时的记录，有 1,500 个类别和物种标签。

BirdClef 2018 is an avian soundscape dataset based on crowd-sourced contributions from the Xeno-canto network. Its training set includes 36,496 audio recordings, covering 1,500 bird species across Central and South America, and it is the largest bioacoustic dataset reported in academic literature. In total, the dataset contains approximately 68 hours of recordings with 1,500 category and species-level labels.

搜集汇总

背景与挑战

背景概述

BirdCLEF 2018是一个大型生物声学数据集，包含36,496个鸟类录音，覆盖1,500种物种，总时长约68小时，主要用于声音事件检测研究。该数据集基于Xeno-canto网络贡献，由多所大学联合发布。