Transformer-based ensemble method for multiple predominant instruments recognition in polyphonic music

EURASIP Journal on Audio, Speech, and Music Processing

Table 3 Precision (P), recall (R), and F1 score for the Swin-T experiments and Han’s model with data augmentation

Class	Han’s Model			Mel-spectrogram			Modgdgram			Tempogram			Voting
	P	R	F1	P	R	F1	P	R	F1	P	R	F1	P	R	F1
Cel	0.55	0.55	0.55	0.52	0.58	0.55	0.27	0.40	0.32	0.52	0.46	0.49	0.61	0.62	0.61
Cla	0.11	0.65	0.18	0.47	0.76	0.58	0.24	0.50	0.33	0.44	0.79	0.56	0.36	0.77	0.49
Flu	0.33	0.61	0.43	0.81	0.83	0.82	0.52	0.63	0.57	0.81	0.80	0.80	0.57	0.80	0.66
Gac	0.84	0.63	0.72	0.43	0.62	0.51	0.64	0.47	0.54	0.30	0.64	0.41	0.59	0.60	0.59
Gel	0.69	0.69	0.69	0.70	0.52	0.60	0.57	0.55	0.56	0.78	0.42	0.55	0.73	0.59	0.66
Org	0.45	0.46	0.45	0.59	0.53	0.56	0.44	0.55	0.49	0.67	0.53	0.59	0.53	0.58	0.55
Pia	0.76	0.61	0.67	0.61	0.54	0.57	0.71	0.47	0.56	0.51	0.50	0.51	0.81	0.51	0.63
Sax	0.62	0.61	0.61	0.68	0.55	0.61	0.53	0.57	0.55	0.78	0.48	0.59	0.61	0.51	0.56
Tru	0.47	0.42	0.44	0.59	0.68	0.63	0.50	0.72	0.59	0.62	0.66	0.64	0.58	0.74	0.65
Vio	0.41	0.57	0.48	0.53	0.59	0.56	0.40	0.63	0.49	0.56	0.55	0.56	0.59	0.73	0.65
Voice	0.94	0.78	0.85	0.70	0.79	0.75	0.57	0.59	0.58	0.77	0.80	0.78	0.69	0.90	0.78
Macro	0.56	0.60	0.55	0.60	0.63	0.61	0.49	0.55	0.51	0.62	0.60	0.59	0.61	0.67	0.62
Micro	0.64	0.64	0.64	0.62	0.62	0.62	0.54	0.54	0.54	0.58	0.58	0.58	0.66	0.66	0.66