GigaMIDI

该数据集名为GigaMIDI，包含了超过140万个独特的MIDI文件，总计18亿个MIDI音符事件以及超过530万个MIDI音轨。它是目前可供研究使用的最大MIDI格式符号音乐集合，特别设计用来区分非表现性MIDI音轨和表现性MIDI音轨。此外，该数据集还包括了每个音轨的循环检测，并经过去重清洗处理，通过Hugging Face Hub提高了可访问性。它由三个主要子集组成：“包含鼓点的所有乐器”、“仅鼓点”和“无鼓点”。具体规模为：1,437,304个独特MIDI文件，5,334,388个MIDI乐器音轨，以及1,824,536,824个MIDI音符事件。该数据集的任务是进行表现性音乐表演检测。The dataset named GigaMIDI contains over 1.4 million unique MIDI files, totaling 1.8 billion MIDI note events and more than 5.3 million MIDI tracks. It is currently the largest collection of MIDI-format notated music available for research, and is specifically designed to distinguish between non-expressive and expressive MIDI tracks. Additionally, the dataset includes cycle detection for each track, has undergone deduplication cleaning, and is hosted on the Hugging Face Hub to enhance accessibility. It consists of three main subsets: "All Instruments with Drums", "Drums Only", and "No Drums". Its specific scale is: 1,437,304 unique MIDI files, 5,334,388 MIDI instrument tracks, and 1,824,536,824 MIDI note events. The task supported by this dataset is expressive musical performance detection.

搜集汇总

背景与挑战

背景概述

GigaMIDI是当前最大的公开MIDI格式符号音乐数据集，包含超过140万个独特MIDI文件、18亿个MIDI音符事件和530万个音轨。它专门区分了非表现性与表现性MIDI音轨，并经过循环检测和去重清洗处理，通过Hugging Face Hub提供访问。该数据集由三个子集（包含鼓点、仅鼓点、无鼓点）组成，主要用于表现性音乐表演检测任务。