MAESTRO v3
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/yonghyunk1m/TowardsRobustTranscription
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为MAESTRO v3，主要用于训练和评估自动钢琴转录模型，特别是在噪声环境中，通过生成加入白噪声的数据集版本。此外，该数据集还用于评估模型在不同信噪比（SNR）下的鲁棒性，信噪比范围从-6分贝到45分贝。测试集包含了177个录音。该数据集的任务是自动钢琴转录。
The dataset named MAESTRO v3 is primarily utilized for training and evaluating automatic piano transcription models, particularly in noisy environments, by generating dataset variants with added white noise. Additionally, it is employed to assess model robustness under different signal-to-noise ratio (SNR) levels ranging from -6 decibels to 45 decibels. The test set consists of 177 recordings. The task associated with this dataset is automatic piano transcription.
提供机构：
MAESTRO
搜集汇总
数据集介绍
以上内容由遇见数据集搜集并总结生成
