def __init__(self, mask: ChunkGrid, verbose=False):
self.mask: ChunkGrid[b8] = mask.copy(dtype=b8)
self.mask.cleanup(remove=True).pad_chunks(1)
self.verbose = verbose
self.__mask_chunks_get = self.mask.chunks.get
self.min, self.max = self.mask.chunks.minmax(True)
self.min -= 1
self.max += 1
def _test_operator2_int(self, op, b0=0, dtype: Type = int, inplace=False):
a = ChunkGrid(2, dtype, 0)
b = ChunkGrid(2, dtype, 1)
a.set_value((0, 0, 0), 1)
a.set_value((0, 0, 1), 1)
b.set_value((0, 0, 0), 1 + b0)
b.set_value((0, 1, 0), 0 + b0)
b.set_value((0, 1, 1), 0 + b0)
a0 = a.copy()
b0 = b.copy()
expected = op(a.to_dense(), b.to_dense())
res = op(a, b)
self.assertIsInstance(res, ChunkGrid)
result = res.to_dense()
assert result.shape == expected.shape
self.assertTrue(np.all(result == expected), f"Failure {op}! \n{result}\n-------\n{expected}")
self._test_inplace_modified(op, a, a0, b, b0, inplace)
def diffuse(model: ChunkGrid[bool], repeat=1):
"""
Diffuse the voxels in model to their neighboring voxels
:param model: the model to diffuse
:param repeat: number of diffusion steps
:return: diffused model
"""
kernel = np.zeros((3, 3, 3), dtype=float)
kernel[1] = 1
kernel[:, 1] = 1
kernel[:, :, 1] = 1
kernel /= np.sum(kernel)
result = ChunkGrid(model.chunk_size, dtype=float, fill_value=1.0)
result[model] = 0.0
result.pad_chunks(repeat // result.chunk_size + 1)
for r in range(repeat):
tmp = result.copy(empty=True)
for chunk in result.chunks:
padded = chunk.padding(result, 1)
ndimage.convolve(padded,
kernel,
output=padded,
mode='constant',
cval=1.0)
conv = padded[1:-1, 1:-1, 1:-1]
m = model.ensure_chunk_at_index(chunk.index, insert=False)
if m.is_filled():
if m.value:
tmp.ensure_chunk_at_index(chunk.index).set_fill(0.0)
continue
else:
conv[m.to_array()] = 0.0
tmp.ensure_chunk_at_index(chunk.index).set_array(conv)
# Expand chunks
for f, i in ChunkGrid.iter_neighbors_indices(chunk.index):
tmp.ensure_chunk_at_index(i)
result = tmp
result.cleanup(remove=True)
return result