Dilated Conv Output Size

Problem

Compute the output size of a dilated convolution.

The effective kernel is dilation * (k - 1) + 1; out = (n + 2 * pad - effective_k) // stride + 1 for each dimension. Return [out_h, out_w].

case 1

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

expected: [26,26]

case 2

in: [28,28,3,1,0,2]

expected: [24,24]

case 3

in: [28,28,3,1,2,2]

expected: [28,28]

case 4

in: [16,16,3,2,1,2]

expected: [7,7]

Starter codePython

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

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