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

Table 3

Test accuracy (%) on different models on CIFAR-10.

Model	Para (M)	GFloat	CIFAR-10
DenseNet-100	7.084	1.875	95.66
DenseNet-100+SE	7.113	1.875	95.74
DenseNet-100+nonlocal	7.084	1.875	95.90
DenseNet-100+CBAM	7.121	1.879	94.53
DenseNet-100+GC	7.117	1.875	95.93
DenseNet-100+LCM	7.100	1.875	96.01
ResNet-50	0.756	0.113	94.14
ResNet-50+SE	0.778	0.113	94.34
ResNet-50+CBAM	0.781	0.114	94.25
ResNet-50+nonlocal	0.758	0.114	94.18
ResNet-50+GC	0.778	0.113	94.10
ResNet-50+LCM	0.778	0.113	94.64