DESED-FL 和 URBAN-FL

DESED-FL和URBAN-FL是专为家庭和城市环境中的声音事件检测（SED）设计的联邦学习（FL）数据集。这些数据集包含大量的音频数据，旨在解决大规模数据处理中的隐私和物流问题。数据集通过模拟真实世界中分布式学习条件，将声音场景分配给具有不同背景特征和类别分布的客户端。这些数据集不仅支持SED的研究，还特别关注FL在处理非独立同分布（non-IID）数据时的挑战，为深度学习模型提供了丰富的训练资源。

DESED-FL and URBAN-FL are federated learning (FL) datasets specifically designed for sound event detection (SED) in household and urban environments. These datasets contain a large volume of audio data, aiming to address the privacy and logistics issues in large-scale data processing. By simulating real-world distributed learning conditions, the datasets assign sound scenes to clients with varying background characteristics and class distributions. These datasets not only support SED research but also specifically focus on the challenges faced by FL when handling non-independent and identically distributed (non-IID) data, providing abundant training resources for deep learning models.