ViT Patch Grid

Problem

Compute the patch grid a Vision Transformer sees for an image.

Return [rows, cols] where rows = h // patch and cols = w // patch. Both image dimensions are divisible by patch.

case 1

in: [224,224,16]

expected: [14,14]

case 2

in: [224,224,32]

expected: [7,7]

case 3

in: [64,96,16]

expected: [4,6]

case 4

in: [32,32,8]

expected: [4,4]

Starter codePython

def patch_grid(h, w, patch):
    # Your code here
    pass

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