-
Notifications
You must be signed in to change notification settings - Fork 41
/
test_formatters_image.py
206 lines (154 loc) · 5.65 KB
/
test_formatters_image.py
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
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
# -*- coding: utf-8 -*-
import pytest
PIL = pytest.importorskip('PIL')
matplotlib = pytest.importorskip('matplotlib')
import numpy as np
from eli5.base import Explanation, TargetExplanation
from eli5.formatters.image import (
format_as_image,
heatmap_to_image,
expand_heatmap,
_validate_heatmap,
_update_alpha,
_cap_alpha,
_overlay_heatmap,
)
from .utils_image import assert_pixel_by_pixel_equal
# 'png' format is required for RGBA data
@pytest.fixture(scope='module')
def boxl():
return PIL.Image.open('tests/images/box_5x5_l.png')
@pytest.fixture(scope='module')
def boxrgb():
return PIL.Image.open('tests/images/box_5x5_rgb.png')
@pytest.fixture(scope='module')
def boxrgba():
return PIL.Image.open('tests/images/box_5x5_rgba.png')
# this is the original catdog image in 'jpg' format with RGB data
@pytest.fixture(scope='module')
def catdog():
return PIL.Image.open('tests/images/cat_dog.jpg')
@pytest.fixture(scope='module')
def catdog_rgba(catdog):
return catdog.convert('RGBA')
def test_validate_heatmap(boxl):
# wrong type
with pytest.raises(TypeError):
_validate_heatmap(boxl)
# out of lower bound
with pytest.raises(ValueError):
_validate_heatmap(np.array([-0.001]))
# out of upper bound
with pytest.raises(ValueError):
_validate_heatmap(np.array([1.001]))
@pytest.mark.parametrize('heatmap', [
(np.zeros((5, 5))),
])
def test_heatmap_to_image_grayscale(heatmap, boxl):
gray_heatmap = heatmap_to_image(heatmap)
assert heatmap.shape == (gray_heatmap.width, gray_heatmap.height)
assert_pixel_by_pixel_equal(gray_heatmap, boxl)
@pytest.mark.parametrize('heatmap', [
(np.zeros((5, 5, 3))),
])
def test_heatmap_to_image_rgb(heatmap, boxrgb):
rgba_heatmap = heatmap_to_image(heatmap)
assert heatmap.shape[:2] == (rgba_heatmap.width, rgba_heatmap.height)
assert_pixel_by_pixel_equal(rgba_heatmap, boxrgb)
@pytest.mark.parametrize('heatmap', [
(np.zeros((5, 5, 4))),
])
def test_heatmap_to_image_rgba(heatmap, boxrgba):
rgba_heatmap = heatmap_to_image(heatmap)
assert heatmap.shape[:2] == (rgba_heatmap.width, rgba_heatmap.height)
assert_pixel_by_pixel_equal(rgba_heatmap, boxrgba)
def test_heatmap_to_image_invalid():
# heatmap must have rank 2 or rank 3
with pytest.raises(ValueError):
heatmap_to_image(np.zeros((1,)))
# coloured heatmap must have 4 or 3 channels
with pytest.raises(ValueError):
heatmap_to_image(np.zeros((1, 1, 10)))
@pytest.mark.parametrize('heatmap, colormap', [
(np.ones((1, 1)), matplotlib.cm.binary),
])
def test_colorize(heatmap, colormap):
colorized = colormap(heatmap)
# check rank
assert len(colorized.shape) == 3
# check that in interval [0, 1]
assert colorized.max() <= 1.0
assert 0.0 <= colorized.min()
@pytest.mark.parametrize('old_arr, alpha_start_arr, new_arr', [
(np.ones((2, 2, 4)), None, np.ones((2, 2, 4))),
(np.zeros((1, 1, 4)), np.ones((1, 1)), np.array([[[0, 0, 0, 1]]])),
])
def test_update_alpha(old_arr, alpha_start_arr, new_arr):
_update_alpha(old_arr, starting_array=alpha_start_arr) # this operation is in-place
assert np.array_equal(old_arr, new_arr)
@pytest.mark.parametrize('alpha_arr, alpha_limit, new_alpha_arr', [
(np.zeros((4, 3)), 0, np.zeros((4, 3))),
(np.array([[0.5, 0.49], [0.51, 0.5]]), 0.5, np.array([[0.5, 0.49], [0.5, 0.5]])),
])
def test_cap_alpha(alpha_arr, alpha_limit, new_alpha_arr):
capped = _cap_alpha(alpha_arr, alpha_limit)
assert np.array_equal(capped, new_alpha_arr)
def test_cap_alpha_invalid():
alpha = np.zeros((1, 1))
# alpha must be a float or int
with pytest.raises(TypeError):
_cap_alpha(alpha, '0.5')
# alpha must be between 0 and 1
with pytest.raises(ValueError):
_cap_alpha(alpha, 1.1)
with pytest.raises(ValueError):
_cap_alpha(alpha, -0.1)
@pytest.mark.parametrize('heatmap', [
(np.zeros((3, 3))),
(np.zeros((10, 10, 4))), # would need downsizing
])
def test_expand_heatmap(boxrgb, heatmap):
expanded = expand_heatmap(heatmap, boxrgb, PIL.Image.BOX)
assert (expanded.width, expanded.height) == (boxrgb.width, boxrgb.height)
def test_expand_heatmap_invalid():
# image is wrong type
heatmap = np.zeros((1, 1))
image = np.ones((2, 2))
with pytest.raises(TypeError):
expand_heatmap(heatmap, image, PIL.Image.BOX)
def test_overlay_heatmap(boxrgba):
overlay = _overlay_heatmap(boxrgba, boxrgba)
assert_pixel_by_pixel_equal(overlay, boxrgba)
@pytest.fixture(scope='module')
def mock_expl(catdog_rgba):
return Explanation('mock estimator',
image=catdog_rgba,
targets=[TargetExplanation(-1,
heatmap=np.zeros((7, 7))
)])
@pytest.fixture(scope='module')
def mock_expl_noheatmap(catdog_rgba):
return Explanation('mock estimator',
image=catdog_rgba,
)
@pytest.fixture(scope='module')
def mock_expl_imgarr():
return Explanation('mock estimator',
image=np.zeros((2, 2, 4)),
)
@pytest.fixture(scope='module')
def mock_expl_imgmode(boxl):
return Explanation('mock estimator',
image=boxl, # mode 'L'
)
def test_format_as_image_notransparency(catdog_rgba, mock_expl):
# heatmap with full transparency
overlay = format_as_image(mock_expl, alpha_limit=0.0)
assert_pixel_by_pixel_equal(overlay, catdog_rgba)
def test_format_as_image_noheatmap(catdog_rgba, mock_expl_noheatmap):
# no heatmap
overlay = format_as_image(mock_expl_noheatmap)
assert_pixel_by_pixel_equal(overlay, catdog_rgba)
def test_format_as_image_invalid_expl(mock_expl_imgarr):
with pytest.raises(TypeError):
format_as_image(mock_expl_imgarr)