DadaGP

DadaGP是一个包含26,181首GuitarPro格式歌曲的数据集，涵盖739种音乐风格。该数据集特别适合用于生成序列模型，如Transformer，其tokenized格式受MIDI编码启发。数据集提供了一个编码/解码工具，用于将GuitarPro文件转换为token格式。DadaGP的应用领域包括吉他-贝斯转录、音乐风格转移和艺术家/风格分类，旨在推动音乐信息检索领域的研究，特别是符号音乐生成任务。DadaGP is a dataset comprising 26,181 GuitarPro-format songs spanning 739 musical genres. This dataset is particularly suitable for sequence generation models such as Transformer, whose tokenized format is inspired by MIDI encoding. The dataset provides an encoding/decoding tool for converting GuitarPro files into tokenized format. Application scenarios of DadaGP include guitar-bass transcription, musical style transfer, and artist/style classification, aiming to advance research in the field of Music Information Retrieval (MIR), especially symbolic music generation tasks.

搜集汇总

背景与挑战

背景概述

DadaGP是一个包含26,181首GuitarPro歌曲的数据集，涵盖739种音乐流派，提供编码器/解码器工具，支持GuitarPro文件与令牌序列格式的相互转换，适用于生成式语言模型。