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

Table 2

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

Model	Para (M)	GFloat	CIFAR-10	CIFAR-100
ResNet-20	0.270	0.041	92.26	68.34
ResNet-20+LCM1	0.278	0.041	92.77	69.40
ResNet-20+LCM2	0.278	0.041	92.70	69.53
ResNet-32	0.464	0.070	93.30	70.67
ResNet-32+LCM1	0.478	0.070	93.90	70.88
ResNet-32+LCM2	0.478	0.070	94.12	71.75