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

Table 1

Test accuracy (%) on CIFAR-10/CIFAR-100 by DenseNet.

Model	Para (M)	GFloat	CIFAR-10	CIFAR-100
DenseNet-40	1.059	0.293	94.67	74.69
DenseNet-40+LCM1	1.065	0.293	95.26	75.61
DenseNet-40+LCM2	1.065	0.293	95.32	75.79
DenseNet-64	2.830	0.761	95.20	77.52
DenseNet-64+LCM1	2.840	0.761	95.84	78.22
DenseNet-64+LCM2	2.840	0.761	95.84	78.24
DenseNet-100	7.084	1.875	95.66	78.76
DenseNet-100+LCM1	7.100	1.875	95.97	79.12
DenseNet-100+LCM2	7.100	1.875	96.01	79.44