Convolutional neural networks: how many pixels will each filter cover?

Question

How can I calculate the area (in the source image) covered by each of the filters on my network?

eg. Assume that the image size is WxW pixels. I am using the following network:

layer 1 : conv :  5x5
layer 2 : pool :  3x3
layer 3 : conv :  5x5
.....
layer N : conv :  5x5

I want to calculate how much area in the original image will be covered by each filter.

eg. the filter in layer 1 will cover 5 × 5 pixels in the original image.

Answer 1

, ? , , ?

, k s, x*x, (((x-k1+1)/s1-k2+1)/s2.../sn)=1, x .

, , , ?

, x/sn,

x = layer[n].k
from i = n-1 to 1
   x = x*layer[i].s + layer[i].k - 1

x*x.

sum_1d - 5, - 5 * 1 + 3-1 = 7, - 5 * 3 + 2 + 4 = 21 ( , s = 3)..

, , , 21 * 21, 17 * 17, (17-2)/3 = 5 ( 16 * 16 15 * 15 ), .