MIMuT Data v2 is a consolidated collection of audio, symbolic music, and note-event annotations assembled for the training and evaluation workflows of MIMuT: Reliable Multi-Instrument Music Transcription, a Transformer-based model for multi-instrument automatic music transcription (AMT).
This repository contains 13 datasets, 82,046 files, and approximately 151.08 GB (decimal; 140.71 GiB) under data/. The collection mixes real recordings, rendered multitracks, and… See the full description on the dataset page:
https://huggingface.co/datasets/J1mmymm/MIMuT_Data_v2.
