Processed Score Graph Dataset for the Score Autoencoder (score-ae)

This is a processed version of the  PDMX dataset (deduplicated and no license conflict) using GraphMuse Library.

This dataset is used by the score-ae repository:https://github.com/hendrik-roth/score-ae

Each graph represents a single music score with up to 1,530 notes and 7 edge types (onset, consecutive, during, rest, and their reverses). Notes carry 14 discrete features: grid position, micro offset, measure index, voice, staff, pitch (step/alter/octave), duration, clef, time signature (beats/beat type), key signature (fifths/mode).

Includes pre-built vocabularies (duration, grid, micro offset) and train/val/test splits (54,833 / 6,854 / 6,855).

For further information on PDMX, see their dataset and paper: Long, Phillip, et al. “PDMX: A Large-Scale Public Domain MusicXML Dataset for Symbolic Music Processing.” ICASSP 2025 – 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, Apr. 2025, pp. 1–5. DOI: 10.1109/ICASSP49660.2025.10890217.For further information on GraphMuse, see their github and paper:E. Karystinaios and G. Widmer, “GraphMuse: A Library for Symbolic Music Graph Processing,” in Proceedings of the International Society for Music Information Retrieval Conference (ISMIR), 2024. Available: https://arxiv.org/abs/2407.12671https://github.com/manoskary/graphmuse/