HRNet Fusion Step

Problem

Fuse a low-resolution branch into a high-resolution branch. Upsample the low-resolution vector by repeating each element (len(high_res) / len(low_res) times), then return high_res + weight * upsampled.

The signature is hrnet_fusion_step(low_res, high_res, weight). The lengths must divide evenly.

Hint

HRNet keeps a high-resolution stream alive and fuses lower-resolution streams into it repeatedly.

Test cases

case 1

in: [[1,2],[10,20,30,40],0.5]

expected: [10.5,20.5,31,41]

case 2

in: [[0,0],[1,1,1,1],1]

expected: [1,1,1,1]

case 3

in: [[5],[2,4],0.5]

expected: [4.5,6.5]

case 4

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

expected: [1,1,2,2,3,3]

Starter codePython

def hrnet_fusion_step(low_res, high_res, weight):
    # Your code here
    pass

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