Tubelet Embedding Parameters

Problem

Count the parameters of a 3D tubelet embedding layer.

A tubelet convolution uses in_ch * dim kernels of shape tube_t x patch x patch plus one bias per output channel. Return in_ch * dim * tube_t * patch * patch + dim.

Hint

One kernel spans all input channels and produces one output channel.

Test cases

case 1

in: [3,768,2,16]

expected: 1180416

case 2

in: [1,192,1,8]

expected: 12480

case 3

in: [3,384,2,14]

expected: 451968

case 4

in: [2,64,4,4]

expected: 8256

Starter codePython

def tubelet_embed_params(in_ch, dim, tube_t, patch):
    # Your code here
    pass

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