A Multifeature Complementary Attention Mechanism for Image Topic Representation in Social Networks

Table 2

The MAP results of CATR and other comparison methods on Sina Weibo image dataset.

Method	MAP@5	MAP@10	MAP@15	MAP@20
mmLDA	0.496	0.527	0.514	0.482
M^3R	0.561	0.582	0.578	0.562
VELDA	0.624	0.642	0.603	0.591
mmETM	0.703	0.734	0.671	0.627
VGG-19	0.732	0.759	0.735	0.706
ResNet	0.739	0.781	0.738	0.717
DenseNet	0.775	0.783	0.741	0.739
SkeletonNet	0.776	0.786	0.745	0.739
CIN	0.779	0.792	0.747	0.741
CATR	0.781	0.801	0.752	0.743