Mural classification model based on high- and low-level vision fusion

Heritage Science

Table 5 Classification performance of the model proposed in this study compared with traditional algorithms

Category	Index	BP	HOG	SIFT	LBP	COLOR	CL	TFNet
Buddha	Precision	56.44	39.66	47.14	51.61	42.65	62.00	72.22
	Recall rate	55.23	37.19	46.52	48.49	40.92	61.06	70.18
	F1-score	55.83	38.39	46.83	50.00	41.77	61.53	71.19
Bodhisattva	Precision	56.52	37.61	43.87	49.78	39.37	37.25	81.44
	Recall rate	55.01	35.62	41.99	48.03	37.67	35.08	80.04
	F1-score	55.75	36.59	42.91	48.89	38.50	36.13	80.73
Disciple	Precision	50.69	27.63	31.93	42.98	40.57	38.20	45.90
	Recall rate	48.69	27.00	30.46	40.91	37.68	37.01	45.01
	F1-score	49.67	27.31	31.18	41.92	39.07	37.60	45.45
Secular person	Precision	63.16	25.45	42.90	25.97	39.80	50.90	91.14
	Recall rate	61.54	23.69	42.11	23.92	37.56	49.09	89.57
	F1-score	62.34	24.54	42.50	24.90	38.65	49.98	90.35
Animal	Precision	49.31	47.35	26.36	45.63	37.08	40.49	70.97
	Recall rate	48.92	45.66	25.10	43.99	35.72	38.88	69.15
	F1-score	49.11	46.49	25.71	44.79	36.39	39.67	70.05
Plant	Precision	60.57	34.68	30.88	47.01	29.25	40.23	61.25
	Recall rate	58.46	32.17	28.55	46.03	28.11	38.93	60.98
	F1-score	59.50	33.38	29.67	46.51	28.67	39.57	61.11
Building	Precision	56.31	11.65	35.81	42.78	31.80	71.11	83.33
	Recall rate	55.22	10.01	33.56	39.86	29.53	67.99	82.83
	F1-score	55.76	10.77	34.65	41.27	30.62	69.52	83.08
Auspicious cloud	Precision	41.41	43.10	29.17	47.26	27.81	30.25	90.91
	Recall rate	40.03	42.14	27.60	45.62	26.59	28.96	90.04
	F1-score	40.71	42.61	28.36	46.43	27.19	29.59	90.47
Average	Precision	54.61	33.68	35.96	43.66	36.04	45.74	80.64
	Recall rate	53.29	31.84	34.06	41.59	35.13	43.52	78.06
	F1-score	53.94	32.73	34.98	42.60	35.58	44.60	78.63

The maximum value of precision rate, recall rate and F1-score for each class of all models are in italics
CL: fused vector of color and texture features