DCASE 2020 Task 1a Development Dataset

该数据集包含了大约14,000个时长为10秒的单通道训练音频剪辑，以及大约3,000个测试音频剪辑，这些剪辑是由9种不同的设备录制的。在训练集中只包含了设备A、B、C以及s1-s3的录音；而设备s4-s6在训练阶段不可用。该数据集包含了使用不同设备录制的音频剪辑，以评估在声学场景分类（ASC）任务中的设备鲁棒性。规模上，训练剪辑有14,000个，测试剪辑有3,000个，任务是指声学场景分类。

This dataset contains approximately 14,000 10-second single-channel training audio clips and around 3,000 test audio clips, all recorded by 9 distinct devices. The training set only includes audio recordings from devices A, B, C, s1, s2 and s3, while devices s4, s5 and s6 are unavailable during the training phase. This dataset features audio clips recorded with different devices, aiming to evaluate device robustness in the Acoustic Scene Classification (ASC) task. In terms of scale, the dataset has 14,000 training clips and 3,000 test clips for the ASC task.