Convolutional Map output dimension

Question

In this paper Let there be Color! implementing CNN, the properties are as follow:

• Image size: 224x224
• Kernel size: 3X3
• Stride: 2X2
• Padding: 1x1

In the paper, they mention that the output layer size is 112X112 but calculating it using the formula Num of outputs = ((Size-F+2*P)/S)+1) [1], the result = 112.5

Shouldn't the output be integer?

Answer 1

I communicated with the paper author and he said

it is implementation-wise, fractional sizes get rounded down so the output would be 112x112

which means the last column will be neglected.