Image Annotation via Reconstitution Graph Learning Model

Table 2

The performance of our proposed method is compared with the previous work on Corel 5K and IAPR TC12 datasets in detail. : average precision; : average recall; : the combination of and ; : number of labels with nonzero recall value.

Method	Corel 5K				IAPR TC12
	(%)	(%)	(%)		(%)	(%)	(%)
CRM [20]	16	19	17	107
MBRM [34]	24	24	24	122	24	23		223
SML [23]	23	29	26	137
JEC [35]	27	32	29	139	29	28		250
TGLM [25]	25	29	27	131
TagProp σSD [27]	28	35	31	145	41	30		259
TagProp ML [27]	31	37	34	146	48	25		227
TagProp σML [27]	33	42	37	160	46	35		266
KSVM-VT [33]	32	42	36	179	47	26		268
FastTag [36]	32	43	37	166	47	26		280
GLKNN [26]	36	47	41	184	41	36		282
2PKNN [28]	39	40	39	177	49	32		274
LDMKL [37]	29	44	35	179
IDFRW [38]	38	49	43	185	49	31	38	275
CCAKNN [29]	41	43	42	185	41	40	41	278
RDPGKNN (this work)	40	45	40	195	40	38	38	283