jech2/lmd-dedup-supplements
收藏官方服务：
资源简介：
LMD去重补充数据集包含从Lakh MIDI数据集（LMD-clean和LMD-full）中使用CAugBERT和CLaMP-1024模型预先计算的嵌入文件。这些嵌入用于去重检测，每个文件夹包括embeddings.pt和refs.txt文件，其中embeddings.pt是嵌入的Torch张量，refs.txt包含与每个嵌入行对应的MIDI文件名。该数据集用于MIDI文件去重的研究，并关联到在ISMIR 2025上发表的论文。
The LMD Deduplication Supplements dataset includes pre-computed embedding files from the Lakh MIDI Dataset (LMD-clean and LMD-full) using the CAugBERT and CLaMP-1024 models for duplicate detection. Each folder contains embeddings.pt, a torch tensor of embeddings, and refs.txt, a list of MIDI filenames corresponding to each embedding row. The dataset is used for research on the deduplication of MIDI files and is associated with a paper presented at ISMIR 2025.
提供机构：
jech2
