This is a strided convolution, where the number of parameters that two convolution computations share during calculation is reduced.
rsvaidya
Strided Convolution also helps to reduce the size of the input so that it becomes feasible to have fully connected layer later where the number of inputs is small.
hweetvpu
Is the line output[(j/STRIDE)*WIDTH...] falsely indented? Shouldn't it be at the same level as the third for loop?
This is a strided convolution, where the number of parameters that two convolution computations share during calculation is reduced.
Strided Convolution also helps to reduce the size of the input so that it becomes feasible to have fully connected layer later where the number of inputs is small.
Is the line
output[(j/STRIDE)*WIDTH...]
falsely indented? Shouldn't it be at the same level as the third for loop?