MIMII DG
收藏资源简介:
MIMII DG数据集是由日立制作所研究开发集团创建,专注于工业机器异常声音检测的领域泛化任务。该数据集包含五种机器类型,每种机器类型对应三种领域转移场景,总计990条数据。数据集通过模拟不同参数值引起的领域转移和难以检测的背景噪音变化,旨在解决实际应用中难以预知的领域转移问题。创建过程中,研究人员通过损坏机器或错误操作机器来录制异常声音,并使用TAMAGO-03麦克风进行录音。该数据集适用于评估和改进异常声音检测系统在未知领域转移条件下的性能。
The MIMII DG Dataset was developed by the Research and Development Group of Hitachi, Ltd., focusing on the domain generalization task for industrial machine anomalous sound detection. This dataset includes five machine types, with three domain shift scenarios corresponding to each machine type, totaling 990 samples. It simulates domain shifts caused by different parameter values and hard-to-detect background noise variations, aiming to solve the unpredictable domain shift issues in real-world applications. During the dataset creation process, researchers recorded anomalous sounds by damaging the machines or operating them incorrectly, using TAMAGO-03 microphones for audio capture. This dataset is applicable to evaluating and improving the performance of anomalous sound detection systems under unknown domain shift conditions.




