The Musical Ternary Modalities Dataset (MTM-Dataset), a public and easily accessible dataset available for learning joint embeddings on musical three modalities for a task in terms of three groups of cross-modal retrieval.
Download: https://drive.google.com/drive/folders/17YuJ07saAQqVgCBeBpk3emrm7X6VX5aV?usp=sharing
The dataset is available for anyone, please cite the following paper:
[1] Dongzeng Zeng, Yi Yu, Keizo Oyama, MTM Dataset for Joint Representation Learning among Sheet Music, Lyrics, and Musical Audio. CSMT2020.
If you have any question, please contact with donghuozeng@gmail.com