TAU Audio-Visual Urban Scenes 2021
收藏arXiv2025-09-30 收录
下载链接：
https://dcase.community/challenge2021/task-acoustic-scene-classification
下载链接
链接失效反馈官方服务：
资源简介：
该数据集为声景分类任务设计了一个基线网络，能够对10种不同的声景进行分类。该网络接受大小为（40×500）的对数梅尔频谱图作为输入，该频谱图对应于一篇10秒的音频剪辑。在推理过程中，该网络的规模包含46,246个参数，以及287百万次的乘累加运算（MACs）。其所承担的任务是声景分类。
This dataset presents a baseline network developed for the acoustic scene classification (ASC) task, which can classify 10 distinct acoustic scenes. The network takes log-Mel spectrograms of size (40 × 500) as input, where each spectrogram corresponds to a 10-second audio clip. During inference, this network has 46,246 parameters and 287 million multiply-accumulate operations (MACs). The task undertaken by this network is acoustic scene classification.
提供机构：
Publicly available
