|
1
|
|
|
import pytest |
|
2
|
|
|
|
|
3
|
|
|
from click.testing import CliRunner |
|
4
|
|
|
from artificial_artwork.cli import cli |
|
5
|
|
|
from unittest.mock import patch |
|
6
|
|
|
|
|
7
|
|
|
|
|
8
|
|
|
runner = CliRunner() |
|
9
|
|
|
|
|
10
|
|
|
@pytest.fixture |
|
11
|
|
|
def image_file_names(): |
|
12
|
|
|
return type('Images', (), { |
|
13
|
|
|
'content': 'canoe_water_w300-h225.jpg', |
|
14
|
|
|
'style': 'blue-red_w300-h225.jpg' |
|
15
|
|
|
}) |
|
16
|
|
|
|
|
17
|
|
|
|
|
18
|
|
|
@pytest.fixture |
|
19
|
|
|
def image_manager(image_manager_class): |
|
20
|
|
|
"""Production ImageManager instance.""" |
|
21
|
|
|
import numpy as np |
|
22
|
|
|
from artificial_artwork.image.image_operations import reshape_image, subtract |
|
23
|
|
|
means = np.array([123.68, 116.779, 103.939]).reshape((1,1,1,3)) |
|
24
|
|
|
return image_manager_class([ |
|
25
|
|
|
lambda matrix: reshape_image(matrix, ((1,) + matrix.shape)), |
|
26
|
|
|
lambda matrix: subtract(matrix, means), # input image must have 3 channels! |
|
27
|
|
|
]) |
|
28
|
|
|
|
|
29
|
|
|
|
|
30
|
|
|
@pytest.fixture |
|
31
|
|
|
def max_iterations_adapter_factory_method(): |
|
32
|
|
|
from artificial_artwork.termination_condition_adapter_factory import TerminationConditionAdapterFactory |
|
33
|
|
|
def create_max_iterations_termination_condition_adapter(iterations): |
|
34
|
|
|
return TerminationConditionAdapterFactory.create('max-iterations', iterations) |
|
35
|
|
|
return create_max_iterations_termination_condition_adapter |
|
36
|
|
|
|
|
37
|
|
|
|
|
38
|
|
|
@pytest.fixture |
|
39
|
|
|
def algorithm_parameters_class(): |
|
40
|
|
|
from artificial_artwork.algorithm import AlogirthmParameters |
|
41
|
|
|
return AlogirthmParameters |
|
42
|
|
|
|
|
43
|
|
|
|
|
44
|
|
|
@pytest.fixture |
|
45
|
|
|
def algorithm(algorithm_parameters_class): |
|
46
|
|
|
from artificial_artwork.algorithm import NSTAlgorithm |
|
47
|
|
|
def _create_algorithm(*parameters): |
|
48
|
|
|
return NSTAlgorithm(algorithm_parameters_class(*parameters)) |
|
49
|
|
|
return _create_algorithm |
|
50
|
|
|
|
|
51
|
|
|
|
|
52
|
|
|
@pytest.fixture |
|
53
|
|
|
def create_algorithm(algorithm, tmpdir): |
|
54
|
|
|
def _create_algorithm(image_manager, termination_condition_adapter): |
|
55
|
|
|
return algorithm( |
|
56
|
|
|
image_manager.content_image, |
|
57
|
|
|
image_manager.style_image, |
|
58
|
|
|
termination_condition_adapter, |
|
59
|
|
|
tmpdir |
|
60
|
|
|
) |
|
61
|
|
|
return _create_algorithm |
|
62
|
|
|
|
|
63
|
|
|
|
|
64
|
|
|
@pytest.fixture |
|
65
|
|
|
def create_production_algorithm_runner(): |
|
66
|
|
|
from artificial_artwork.nst_tf_algorithm import NSTAlgorithmRunner |
|
67
|
|
|
from artificial_artwork.image.image_operations import noisy, convert_to_uint8 |
|
68
|
|
|
from artificial_artwork.styling_observer import StylingObserver |
|
69
|
|
|
from artificial_artwork.disk_operations import Disk |
|
70
|
|
|
|
|
71
|
|
|
noisy_ratio = 0.6 |
|
72
|
|
|
def _create_production_algorithm_runner(termination_condition_adapter): |
|
73
|
|
|
algorithm_runner = NSTAlgorithmRunner.default( |
|
74
|
|
|
lambda matrix: noisy(matrix, noisy_ratio), |
|
75
|
|
|
) |
|
76
|
|
|
|
|
77
|
|
|
algorithm_runner.progress_subject.add( |
|
78
|
|
|
termination_condition_adapter, |
|
79
|
|
|
) |
|
80
|
|
|
algorithm_runner.persistance_subject.add( |
|
81
|
|
|
StylingObserver(Disk.save_image, convert_to_uint8) |
|
82
|
|
|
) |
|
83
|
|
|
return algorithm_runner |
|
84
|
|
|
return _create_production_algorithm_runner |
|
85
|
|
|
|
|
86
|
|
|
|
|
87
|
|
|
@pytest.fixture |
|
88
|
|
|
def get_algorithm_runner(create_production_algorithm_runner): |
|
89
|
|
|
def _get_algorithm_runner(termination_condition_adapter): |
|
90
|
|
|
algorithm_runner = create_production_algorithm_runner( |
|
91
|
|
|
termination_condition_adapter, |
|
92
|
|
|
) |
|
93
|
|
|
return algorithm_runner |
|
94
|
|
|
return _get_algorithm_runner |
|
95
|
|
|
|
|
96
|
|
|
|
|
97
|
|
|
@pytest.fixture |
|
98
|
|
|
def get_model_design(): |
|
99
|
|
|
def _get_model_design(handler, network_design): |
|
100
|
|
|
return type('ModelDesign', (), { |
|
101
|
|
|
'pretrained_model': handler, |
|
102
|
|
|
'network_design': network_design}) |
|
103
|
|
|
return _get_model_design |
|
104
|
|
|
|
|
105
|
|
|
|
|
106
|
|
|
def test_nst_runner( |
|
107
|
|
|
get_algorithm_runner, |
|
108
|
|
|
create_algorithm, |
|
109
|
|
|
image_file_names, |
|
110
|
|
|
get_model_design, |
|
111
|
|
|
max_iterations_adapter_factory_method, |
|
112
|
|
|
image_manager, |
|
113
|
|
|
test_image, |
|
114
|
|
|
model, |
|
115
|
|
|
tmpdir): |
|
116
|
|
|
"""Test nst algorithm runner. |
|
117
|
|
|
|
|
118
|
|
|
Runs a simple 'smoke test' by iterating only 3 times. |
|
119
|
|
|
""" |
|
120
|
|
|
import os |
|
121
|
|
|
ITERATIONS = 3 |
|
122
|
|
|
|
|
123
|
|
|
image_manager.load_from_disk(test_image(image_file_names.content), 'content') |
|
124
|
|
|
image_manager.load_from_disk(test_image(image_file_names.style), 'style') |
|
125
|
|
|
|
|
126
|
|
|
assert image_manager.images_compatible == True |
|
127
|
|
|
|
|
128
|
|
|
termination_condition_adapter = max_iterations_adapter_factory_method(ITERATIONS) |
|
129
|
|
|
|
|
130
|
|
|
algorithm_runner = get_algorithm_runner(termination_condition_adapter) |
|
131
|
|
|
|
|
132
|
|
|
algorithm = create_algorithm(image_manager, termination_condition_adapter) |
|
133
|
|
|
|
|
134
|
|
|
model_design = get_model_design( |
|
135
|
|
|
model.pretrained_model.handler, |
|
136
|
|
|
model.network_design, |
|
137
|
|
|
) |
|
138
|
|
|
model_design.pretrained_model.load_model_layers() |
|
139
|
|
|
algorithm_runner.run(algorithm, model_design) |
|
140
|
|
|
|
|
141
|
|
|
template_string = image_file_names.content + '+' + image_file_names.style + '-' + '{}' + '.png' |
|
142
|
|
|
assert os.path.isfile(os.path.join(tmpdir, template_string.format(1))) |
|
143
|
|
|
assert os.path.isfile(os.path.join(tmpdir, template_string.format(ITERATIONS))) |
|
144
|
|
|
|
boromir674 /
neural-style-transfer
