Semi-Supervised Cross-Modal Retrieval Based on Discriminative Comapping

Table 1

Retrieval performance (mAP) on three public datasets.

wiki-R				wiki-W
Methods	I2T (%)	T2I (%)	Average (%)	I2T (%)	T2I (%)	Average (%)
PLS	23.75	17.23	20.49	35.95	35.10	35.53
CCA	24.14	19.71	21.93	33.16	31.66	32.41
SM	22.64	21.84	22.24	36.85	38.67	37.76
SCM	26.62	22.57	24.59	37.48	39.26	38.37
GMMFA	23.09	20.34	21.72	28.41	24.87	26.64
GMLDA	24.64	19.52	22.08	30.03	28.06	29.05
JLSLR	23.60	21.22	22.41	39.42	36.91	38.17
MDCR	26.19	21.03	23.61	41.07	37.75	39.41
ACMR	33.22	24.50	28.86	50.61	42.82	46.72
SGRCR	28.42	22.71	25.57	43.65	40.60	42.10
MRRDC	28.93	26.29	27.61	62.34	53.00	57.67
	Pascal sentence			INRIA-Websearch
PLS	36.53	37.63	37.08	19.38	26.03	22.71
CCA	37.99	37.20	37.59	26.03	27.95	26.99
SM	44.98	43.39	44.19	37.83	35.31	36.57
SCM	40.71	39.35	40.03	35.44	30.87	33.16
GMMFA	37.32	34.70	36.01	28.09	30.37	29.23
GMLDA	40.80	38.77	39.79	47.59	54.07	50.83
JLSLR	45.42	45.56	45.49	52.51	54.53	53.52
MDCR	43.22	46.22	44.72	47.09	45.99	46.54
ACMR	46.81	56.23	51.52	55.85	66.92	61.39
SGRCR	49.23	50.00	49.60	54.10	55.40	54.78
MRRDC	66.49	58.54	62.52	56.70	68.81	62.76