Transposed Conv with Output Padding

Problem

Compute the spatial output size of a transposed convolution including output padding.

out = (n - 1) * stride - 2 * pad + k + output_padding per dimension, where output_padding < stride resolves ambiguous output sizes. Given h, w, kernel size k, stride, pad and output_padding, return [out_h, out_w].

Test cases

case 1

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

expected: [14,14]

case 2

in: [4,4,3,1,0,0]

expected: [6,6]

case 3

in: [5,10,3,2,0,1]

expected: [12,22]

Starter codePython

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

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