Patch Embedding Dimension

Problem

Compute the projection dimension of a convolutional patch embedding.

A patch of patch_h x patch_w pixels from an image with in_channels channels flattens to patch_h * patch_w * in_channels values. Return that dimension.

Test cases

case 1

in: [16,16,3]

expected: 768

case 2

in: [14,14,1]

expected: 196

case 3

in: [8,8,3]

expected: 192

case 4

in: [4,4,64]

expected: 1024

Starter codePython

def patch_embed_dim(patch_h, patch_w, in_channels):
    # Your code here
    pass

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