Music and Audio Research Group (MARG), Department of Intelligence and Information, Seoul National University, Seoul, Republic of Korea
Below are the chord examples that are generated from three proposed models (STHarm, VTHarm, rVTHarm), two baseline models (BLSTM, ONADE), and the ground truth (GT) from Chord Melody Dataset and Hooktheory Lead Sheet Dataset. The examples for Chord Melody Dataset are especially from the actual listening test for the paper.
Song 1
Song 2
Song 3
Melody
melody__anthropology__s0_p8-23.wav
melody__groovn_high__s10_p16-31.wav
melody__fascinating_rhythm__s8_p4-19.wav
GT
GT__anthropology__s0_p8-23.wav
GT__groovn_high__s10_p16-31.wav
GT__fascinating_rhythm__s8_p4-19.wav
BLSTM (2017)
sampled__anthropology__s0_base_bilstm_CMD_100_p8-23.wav
sampled__groovn_high__s10_base_bilstm_CMD_100_p16-31.wav
sampled__fascinating_rhythm__s8_base_bilstm_CMD_100_p4-19.wav
ONADE (2020)
sampled__anthropology__s0_base_nade_CMD_100_p8-23.wav
sampled__groovn_high__s10_base_nade_CMD_100_p16-31.wav
sampled__fascinating_rhythm__s8_base_nade_CMD_100_p4-19.wav
STHarm
sampled__anthropology__s0_base_transformer_l4_256_CMD_100_p8-23.wav
sampled__groovn_high__s10_base_transformer_l4_256_CMD_100_p16-31.wav
sampled__fascinating_rhythm__s8_base_transformer_l4_256_CMD_100_p4-19.wav
VTHarm
sampled__anthropology__s0_exp382_noReg_l4_256_CMD_100_p8-23_c-0.3000.wav
sampled__groovn_high__s10_exp382_noReg_l4_256_CMD_100_p16-31_c0.8000.wav
sampled__fascinating_rhythm__s8_exp382_noReg_l4_256_CMD_100_p4-19_c0.2000.wav
rVTHarm (α=0)
sampled__anthropology__s0_exp382_l4_256_CMD_100_p8-23_c0.0000.wav
sampled__groovn_high__s10_exp382_l4_256_CMD_100_p16-31_c0.0000.wav
sampled__fascinating_rhythm__s8_exp382_l4_256_CMD_100_p4-19_c0.0000.wav