Table 3.

Network architecture details.

No	Name	Type	Activations	Learnable	Total Learnable
1	Image input 28 × 28 × 1 with ‘zerocente…	Image input	28 × 28 × 1	-	0
2	Conv_1 3 5 × 5 × 1 convolutions with stride…	Convolution	28 × 28 × 3	Weights 5 × 5 × 1 × 3 Bias 1 × 1 × 3	78
3	Conv_3 3 5 × 5 × 1 convolutions with stride…	Convolution	28 × 28 × 3	Weights 5 × 5 × 1 × 3 Bias 1 × 1 × 3	78
4	reLu_3 ReLU	ReLU	28 × 28 × 3	-	0
5	batchnorm_3 Batch normalization with 3 chan...	Batch normalization	28 × 28 × 3	Offset 1 × 1 × 3 Scale 1 × 1 × 3	6
6	maxpool_3 5 × 5 max pool with stride [1,1]	Max pooling	28 × 28 × 3	-	0
7	reLu_1 ReLU	ReLU	28 × 28 × 3	-	0
8	batchnorm_1 Batch normalization with 3 chan..	Batch normalization	28 × 28 × 3	Offset 1 × 1 × 3 Scale 1 × 1 × 3	6
9	maxpool_1 5 × 5 max pool with stride [1,1]	Max pooling	28 × 28 × 3	-	0
10	Conv_4 3 5 × 5 × 1 convolutions with stride…	Convolution	28 × 28 × 3	Weights 5 × 5 × 1 × 3 Bias 1 × 1 × 3	78
11	reLu_4, ReLU	ReLU	28 × 28 × 3	-	0
12	batchnorm_4 Batch normalization with 3 chan..	Batch normalization	28 × 28 × 3	Offset 1 × 1 × 3 Scale 1 × 1 × 3	6
13	maxpool_4 5 × 5 max pool with stride [1,1]	Max pooling	28 × 28 × 3	-	0
14	Conv_2 3 5 × 5 × 1 convolutions with stride…	Convolution	28 × 28 × 3	Weights 5 × 5 × 1 × 3 Bias 1 × 1 × 3	78
15	reLu_2, ReLU	ReLU	28 × 28 × 3	-	0
16	batchnorm_2 Batch normalization with 3 chan…	Batch normalization	28 × 28 × 3	Offset 1 × 1 × 3 Scale 1 × 1 × 3	6
17	maxpool_2 5 × 5 max pool with stride [1,1]	Max pooling	28 × 28 × 3	-	0
18	Depthcat Depth concatenation of 4 inputs	Depth concatenation	28 × 28 × 12	-	0
19	Fc 12 fully connected layer	Fully connected	1 × 1 × 12	Weights 12 × 9408 Bias 12 × 1	112,908
20	SoftMax	SoftMax	1 × 1 × 12	-	0
21	Classoutput crossentropyex with ‘index-L’…..	Classification output	-	-	0