Table 1.

Hyperparameters considered in proposed CES-CNN

Block #	Layer	Output shape	Image size	Parameters
Block-1 (Conv-1)	Convolution2D ($3\times 3$)@64	(n,n,64)	(128,128,64)	(($3\times 3$)+1)$\times$64 = 640
	Batch Normalization	(n,n,64)	(128,128,64)	$4\times 64$=256
	Activation Relu	(n,n,64)	(128,128,64)	0
	Maxpooling2D ($2\times 2$)	(n1,n1,64) n1=n/2	(64,64,64)	0
	Dropout	(n1,n1,64)	(64,64,64)	0
Block-2 (Conv-2)	Convolution2D ($3\times 3$)@128	(n1,n1,128)	(64,64,128)	(($3\times 3\times$64)+1) $\times$128=73876
	Batch Normalization	(n1,n1,128)	(64,64,128)	$4\times 128$=512
	Activation Relu	(n1,n1,128)	(64,64,128)	0
	Maxpooling2D ($2\times 2$)	(n2,n2,128) n2=n1/2	(32,32,128)	0
	Dropout	(n2,n2,128)	(32,32,128)	0
Block-3 (Conv-3)	Convolution2D ($3\times 3$)@256	(n2,n2,256)	(32,32,256)	(($3\times 3\times$128)+1) $\times$256=295168
	Batch Normalization	(n2,n2,256)	(32,32,256)	$4\times 256$=1024
	Activation Relu	(n2,n2,128)	(32,32,256)	0
	Maxpooling2D ($2\times 2$)	(n3,n3,256) n3=n2/2	(16,16,256)	0
	Dropout	(n3,n3,256)	(16,16,256)	0
Block-4 (Conv-4)	Convolution2D ($3\times 3$)@512	(n3,n3,512)	(16,16,512)	(($3\times 3\times$256)+1) $\times$512=1180160
	Batch Normalization	(n3,n3,256)	(16,16,512)	$4\times 512$=2048
	Activation Relu	(n3,n3,256)	(16,16,512)	0
	Maxpooling2D ($2\times 2$)	(n4,n4,512) n4=n3/2	(8,8,512)	0
	Dropout	(n4,n4,256)	(8,8,512)	0
Block-5 (Conv-5)	Convolution2D ($3\times 3$)@512	(n4,n4,512)	(8,8,512)	(($3\times 3\times$512)+1) $\times$512=2359808
	Batch Normalization	(n4,n4,512)	(8,8,512)	$4\times 512$=2048
	Activation Relu	(n4,n4,512)	(8,8,512)	0
	Maxpooling2D ($2\times 2$)	(n5,n5,512) n5=n4/2	(4,4,512)	0
	Dropout	(n5,n5,512)	(4,4,512)	0
	Flatten	(1,$n5\times n5\times$512	(1,8192)	0
	Dense	(1256)	(1,256)	(8192+1)$\times$256 = 2097408
	Batch Normalization	(1,256)	(1,256)	512
	activation Relu	(1,256)	(1,256)	0
	Dropout	(1,256)	(1,256)	0
	Dense	(1,512)	(1,512)	(256+1)$\times$512 = 131584
	Batch Normalization	(1,512)	(1,512)	1024
	activation Relu	(1,512)	(1,512)	0
	Dropout	(1,512)	(1,512)	0
	Dense	(1,7)	(1,7)	(512+1)$\times$7 = 3591
	Total parameters for image of size ($128\times 128$)			9481607