Skip to main content

Advertisement

Table 1 MAP scores of our RCN and compared methods

From: Cross-media residual correlation learning

Dataset Method Task
Image\(\rightarrow\)Text Text\(\rightarrow\)Image Average
Wikipedia dataset CCA 0.176 0.178 0.177
CFA 0.330 0.306 0.318
KCCA (Poly) 0.230 0.224 0.227
KCCA (Gaussian) 0.357 0.328 0.343
Bimodal AE 0.301 0.267 0.284
Multimodal DBN 0.204 0.145 0.175
Corr-AE 0.373 0.357 0.365
JRL 0.408 0.353 0.381
LGCFL 0.416 0.360 0.388
CMDN 0.409 0.364 0.387
Deep-SM 0.458 0.345 0.402
RCN (OnlyCorrelation) 0.465 0.407 0.436
Our RCN 0.489 0.418 0.454
NUS-WIDE-10k dataset CCA 0.159 0.189 0.174
CFA 0.299 0.301 0.300
KCCA (Poly) 0.129 0.157 0.143
KCCA (Gaussian) 0.295 0.162 0.229
Bimodal AE 0.234 0.376 0.305
Multimodal DBN 0.178 0.144 0.161
Corr-AE 0.306 0.340 0.323
JRL 0.410 0.444 0.427
LGCFL 0.408 0.374 0.391
CMDN 0.410 0.450 0.430
Deep-SM 0.389 0.496 0.443
RCN (OnlyCorrelation) 0.360 0.406 0.383
Our RCN 0.497 0.517 0.507
Pascal Sentences dataset CCA 0.110 0.116 0.113
CFA 0.341 0.308 0.325
KCCA (Poly) 0.271 0.280 0.276
KCCA (Gaussian) 0.312 0.329 0.321
Bimodal AE 0.404 0.447 0.426
Multimodal DBN 0.438 0.363 0.401
Corr-AE 0.411 0.475 0.443
JRL 0.416 0.377 0.397
LGCFL 0.381 0.435 0.408
CMDN 0.458 0.444 0.451
Deep-SM 0.440 0.414 0.427
RCN (OnlyCorrelation) 0.433 0.443 0.438
Our RCN 0.472 0.453 0.463