Growcut

[marcdk]

Stéfans GrowCut code. I've checked that the number of changes does go down to zero.

[ahmadia]

I've written some tests in https://www.wakari.io/sharing/bundle/aron/Accelerating_Python_Libraries_with_Numba_-_Part_2

You're free to reuse them with attribution.

[ahmadia]

Specifically, this hunk of code:

def test_kernel(kernel):

    # Cython creates a built-in function that is not modifiable and does not possess the func_name attribute
    if hasattr(kernel, 'func_name'):
        name = kernel.func_name
    else:
        name = kernel.__name__ 

    image = np.zeros((3, 3, 3))
    state = np.zeros((3, 3, 2))
    state_next = np.empty_like(state)

    # colony 1 is strength 1 at position 0,0
    # colony 0 is strength 0 at all other positions
    state[0, 0, 0] = 1
    state[0, 0, 1] = 1

    # window_size 1, colony 1 should propagate to three neighbors
    changes = kernel(image, state, state_next, 1)
    assert(3 == changes)
    npt.assert_array_equal(state_next[0:2, 0:2], 1)
    npt.assert_array_equal(state_next[2, :], 0)
    npt.assert_array_equal(state_next[2, :], 0)

    # window_size 2, colony 1 should propagate to entire image
    changes = kernel(image, state, state_next, 2)
    assert(8 == changes)
    npt.assert_array_equal(state_next, 1)

    print "PASS", name

[alexdesiqueira]

Hey @stefanv, are you interested in this one? Thanks!