1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
|
import astra
import astra.experimental
import numpy as np
import pytest
DET_SPACING_X = 1.0
DET_SPACING_Y = 1.0
DET_ROW_COUNT = 20
DET_COL_COUNT = 45
N_ANGLES = 180
ANGLES = np.linspace(0, 2 * np.pi, N_ANGLES, endpoint=False)
N_ROWS = 40
N_COLS = 30
N_SLICES = 50
VOL_GEOM = astra.create_vol_geom(N_ROWS, N_COLS, N_SLICES)
PROJ_GEOM = astra.create_proj_geom('parallel3d', DET_SPACING_X, DET_SPACING_Y,
DET_ROW_COUNT, DET_COL_COUNT, ANGLES)
DATA_INIT_VALUE = 1.0
def _convert_to_backend(data, backend):
if backend == 'numpy':
return data
elif backend == 'pytorch_cpu':
import torch
return torch.tensor(data, device='cpu')
elif backend == 'pytorch_cuda':
import torch
return torch.tensor(data, device='cuda')
elif backend == 'cupy':
import cupy as cp
return cp.array(data)
elif backend == 'jax_cpu':
import jax
return jax.device_put(data, device=jax.devices('cpu')[0])
elif backend == 'jax_cuda':
import jax
return jax.device_put(data, device=jax.devices('cuda')[0])
@pytest.fixture
def projector():
projector_id = astra.create_projector('cuda3d', PROJ_GEOM, VOL_GEOM)
yield projector_id
astra.projector3d.delete(projector_id)
@pytest.fixture
def vol_data(backend):
data = np.full([N_SLICES, N_ROWS, N_COLS], DATA_INIT_VALUE, dtype=np.float32)
return _convert_to_backend(data, backend)
@pytest.fixture
def proj_data(backend):
data = np.full([DET_ROW_COUNT, N_ANGLES, DET_COL_COUNT], DATA_INIT_VALUE, dtype=np.float32)
return _convert_to_backend(data, backend)
@pytest.fixture
def reference_fp():
vol_data_id = astra.data3d.create('-vol', VOL_GEOM, DATA_INIT_VALUE)
data_id, data = astra.create_sino3d_gpu(vol_data_id, PROJ_GEOM, VOL_GEOM)
astra.data3d.delete(data_id)
return data
@pytest.fixture
def reference_bp():
proj_data_id = astra.data3d.create('-sino', PROJ_GEOM, DATA_INIT_VALUE)
data_id, data = astra.create_backprojection3d_gpu(proj_data_id, PROJ_GEOM, VOL_GEOM)
astra.data3d.delete(data_id)
return data
@pytest.fixture
def proj_data_non_contiguous(backend):
data = np.full([N_ANGLES, DET_ROW_COUNT, DET_COL_COUNT], DATA_INIT_VALUE, dtype=np.float32)
return _convert_to_backend(data, backend).swapaxes(0, 1)
@pytest.fixture
def vol_geom_singular_dim(singular_dims):
dims = [N_ROWS, N_COLS, N_SLICES]
for dim in singular_dims.split('-'):
if dim == 'rows':
dims[0] = 1
elif dim == 'cols':
dims[1] = 1
elif dim == 'slices':
dims[2] = 1
return astra.create_vol_geom(*dims)
@pytest.fixture
def vol_data_singular_dim(singular_dims, backend):
shape = [N_SLICES, N_ROWS, N_COLS]
for dim in singular_dims.split('-'):
if dim == 'rows':
shape[1] = 1
elif dim == 'cols':
shape[2] = 1
elif dim == 'slices':
shape[0] = 1
data = np.full(shape, DATA_INIT_VALUE, dtype=np.float32)
return _convert_to_backend(data, backend)
@pytest.fixture
def vol_data_slice(singular_dims, backend):
data = np.full([N_SLICES, N_ROWS, N_COLS], DATA_INIT_VALUE, dtype=np.float32)
data = _convert_to_backend(data, backend)
for dim in singular_dims.split('-'):
if dim == 'rows':
data = data[:, 0:1, :]
elif dim == 'cols':
data = data[:, :, 0:1]
elif dim == 'slices':
data = data[0:1, :, :]
return data
@pytest.mark.parametrize('backend', ['numpy', 'pytorch_cpu', 'pytorch_cuda', 'cupy',
'jax_cpu', 'jax_cuda'])
class TestAll:
def test_backends_fp(self, backend, projector, vol_data, proj_data, reference_fp):
astra.experimental.direct_FP3D(projector, vol_data, proj_data)
if backend.startswith('pytorch'):
proj_data = proj_data.cpu()
assert np.allclose(proj_data, reference_fp)
def test_backends_bp(self, backend, projector, vol_data, proj_data, reference_bp):
astra.experimental.direct_BP3D(projector, vol_data, proj_data)
if backend.startswith('pytorch'):
vol_data = vol_data.cpu()
assert np.allclose(vol_data, reference_bp)
def test_non_contiguous(self, backend, proj_data_non_contiguous):
if backend.startswith('jax'):
# JAX should not produce non-contiguous tensors, so nothing to test
return
with pytest.raises(ValueError):
astra.data3d.link('-sino', PROJ_GEOM, proj_data_non_contiguous)
@pytest.mark.parametrize('singular_dims', [
'rows', 'cols', 'slices', 'rows-cols', 'rows-slices', 'cols-slices', 'rows-cols-slices'
])
def test_singular_dimensions(self, backend, singular_dims, vol_geom_singular_dim,
vol_data_singular_dim):
astra.data3d.link('-vol', vol_geom_singular_dim, vol_data_singular_dim)
@pytest.mark.parametrize('backend', ['pytorch_cuda', 'cupy'])
@pytest.mark.parametrize('singular_dims', ['rows'])
def test_allow_pitched(backend, singular_dims, vol_geom_singular_dim, vol_data_slice):
astra.data3d.link('-vol', vol_geom_singular_dim, vol_data_slice)
@pytest.mark.parametrize('backend', ['numpy'])
def test_read_only(backend, vol_data):
vol_data.flags['WRITEABLE'] = False
# BufferError for numpy < 2 which doesn't support exporting read-only arrays
# ValueError for numpy >= 2 where astra rejects the read-only array
with pytest.raises((ValueError, BufferError)):
astra.data3d.link('-vol', VOL_GEOM, vol_data)
|
