[Retracted] Channel-Wise Correlation Calibrates Attention Module for Convolutional Neural Networks

Table 4

Test accuracy (%) on different models on CIFAR-100.

Model	Para (M)	GFloat	CIFAR-10
DenseNet-100	7.084	1.875	78.76
DenseNet-100+SE	7.113	1.875	79.41
DenseNet-100+nonlocal	7.084	1.875	79.22
DenseNet-100+CBAM	7.121	1.879	75.86
DenseNet-100+GC	7.117	1.875	79.14
DenseNet-100+LCM	7.100	1.875	79.44
ResNet-50	0.756	0.113	72.25
ResNet-50+SE	0.778	0.113	73.04
ResNet-50+CBAM	0.781	0.114	72.95
ResNet-50+nonlocal	0.758	0.114	72.82
ResNet-50+GC	0.778	0.113	71.85
ResNet-50+LCM	0.778	0.113	73.87