Conv Output Size

Problem

Compute the spatial output size of a strided 2D convolution.

out = (n + 2 * pad - k) // stride + 1 for each dimension. Given height h, width w, kernel size k, stride and pad, return [out_h, out_w].

Hint

Padding adds 2 * pad to each side before the kernel slides.

case 1

in: [28,28,3,1,0]

expected: [26,26]

case 2

in: [32,32,3,1,1]

expected: [32,32]

case 3

in: [28,28,5,2,0]

expected: [12,12]

case 4

in: [7,7,3,2,1]

expected: [4,4]

Starter codePython

def conv_output_size(h, w, k, stride, pad):
    # Your code here
    pass

4 test cases · run your code against them in the editor