mcht67/Polyphonic-BirdSet-train
收藏资源简介:
该数据集是一个综合性的鸟类声音音频数据集,包含多个配置变体,主要用于鸟类声音识别、分类和声景分析。数据集包含以下主要配置:1. HSN_noise:包含带噪声的鸟类声音片段,用于测试噪声环境下的识别性能;2. HSN_polyphonic_6和HSN_polyphonic_8:模拟多只鸟类同时发声的多音场景,支持最多6或8个声源,包含噪声和无噪声版本,并提供了多种预训练音频模型的嵌入特征(如BirdNET、VGGish、YAMNet、Perch等);3. HSN_segments:提供鸟类声音的原始片段,用于基础训练和评估;4. HSN_soundscape_test:用于声景测试的鸟类声音数据;5. NES_noise:包含来自NES(可能指北美鸟类声音)的噪声数据。数据集特征包括音频文件、鸟类物种标签(使用eBird代码、属、物种组、目分类)、多标签支持、地理坐标(经纬度)、时间戳、音频技术参数(频率范围、信噪比、增益等)、录音设备信息和许可证信息。数据集已划分为训练集、验证集和测试集,适用于机器学习模型训练和评估,特别是在生物声学和生态监测领域。
This dataset is a comprehensive collection of bird sound audio data with multiple configuration variants, primarily designed for bird sound recognition, classification, and soundscape analysis. The dataset includes the following main configurations: 1. HSN_noise: Contains bird sound clips with noise for testing recognition performance in noisy environments; 2. HSN_polyphonic_6 and HSN_polyphonic_8: Simulate polyphonic scenarios with multiple birds vocalizing simultaneously, supporting up to 6 or 8 sound sources, including noisy and noise-free versions, and providing embeddings from various pre-trained audio models (e.g., BirdNET, VGGish, YAMNet, Perch); 3. HSN_segments: Provides raw bird sound segments for basic training and evaluation; 4. HSN_soundscape_test: Bird sound data for soundscape testing; 5. NES_noise: Includes noise data from NES (likely referring to North American bird sounds). Dataset features include audio files, bird species labels (using eBird codes, genus, species group, order classification), multi-label support, geographic coordinates (latitude and longitude), timestamps, audio technical parameters (frequency range, SNR, gain, etc.), recording device information, and license details. The dataset is split into training, validation, and test sets, making it suitable for machine learning model training and evaluation, particularly in bioacoustics and ecological monitoring applications.





