-
Notifications
You must be signed in to change notification settings - Fork 699
/
test_cleanvision_integration.py
302 lines (253 loc) · 10.7 KB
/
test_cleanvision_integration.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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
import matplotlib.pyplot as plt
import numpy as np
import pytest
import pandas as pd
from cleanlab import Datalab
import cleanlab.datalab.internal.adapter.imagelab as imagelab
LABEL_NAME = "label"
IMAGE_NAME = "image"
IMAGELAB_ISSUE_TYPES = [
"dark",
"light",
"low_information",
"odd_aspect_ratio",
"odd_size",
"grayscale",
"blurry",
]
SEED = 42
class TestCleanvisionIntegration:
@pytest.fixture
def features(self, image_dataset):
np.random.seed(SEED)
return np.random.rand(len(image_dataset), 5)
@pytest.fixture
def num_imagelab_issues(self):
return 7
@pytest.fixture
def num_datalab_issues(self):
return 6
@pytest.fixture
def pred_probs(self, image_dataset):
np.random.seed(SEED)
return np.random.rand(len(image_dataset), 2)
@pytest.fixture
def set_plt_show(self, monkeypatch):
monkeypatch.setattr(plt, "show", lambda: None)
@pytest.mark.usefixtures("set_plt_show")
def test_imagelab_issues_checked(
self, image_dataset, pred_probs, features, capsys, num_imagelab_issues, num_datalab_issues
):
datalab = Datalab(data=image_dataset, label_name=LABEL_NAME, image_key=IMAGE_NAME)
datalab.find_issues(pred_probs=pred_probs, features=features)
captured = capsys.readouterr()
assert (
"Finding dark, light, low_information, odd_aspect_ratio, odd_size, grayscale, blurry images"
in captured.out
)
# unable to check for non iid as feature space is too small, skipping it in interest of time
assert "Failed to check for these issue types: [NonIIDIssueManager]" in captured.out
assert len(datalab.issues) == len(image_dataset)
# add up imagelab + datalab issues
assert len(datalab.issues.columns) == (num_imagelab_issues + num_datalab_issues) * 2
assert len(datalab.issue_summary) == num_imagelab_issues + num_datalab_issues
all_keys = IMAGELAB_ISSUE_TYPES + [
"statistics",
"label",
"outlier",
"near_duplicate",
"class_imbalance",
"null",
"underperforming_group",
# "non_iid",
]
assert set(all_keys) == set(datalab.info.keys())
datalab.report(show_all_issues=True)
captured = capsys.readouterr()
for issue_type in IMAGELAB_ISSUE_TYPES:
assert issue_type in captured.out
df = pd.DataFrame(
{
"issue_type": [
"dark",
"light",
"low_information",
"odd_aspect_ratio",
"odd_size",
"grayscale",
"blurry",
"label",
"outlier",
"near_duplicate",
"class_imbalance",
"null",
"underperforming_group",
],
"num_issues": [1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0],
}
)
expected_count = df.sort_values(by="issue_type")["num_issues"].tolist()
count = datalab.issue_summary.sort_values(by="issue_type")["num_issues"].tolist()
assert set(datalab.issue_summary["issue_type"].tolist()) == set(df["issue_type"].tolist())
assert count == expected_count
assert datalab.issue_summary["num_issues"].sum() == df["num_issues"].sum()
@pytest.mark.usefixtures("set_plt_show")
def test_imagelab_max_prevalence(
self,
image_dataset,
pred_probs,
features,
capsys,
num_datalab_issues,
monkeypatch,
):
max_prevalence = 0
monkeypatch.setattr(imagelab, "IMAGELAB_ISSUES_MAX_PREVALENCE", max_prevalence)
datalab = Datalab(data=image_dataset, label_name=LABEL_NAME, image_key=IMAGE_NAME)
datalab.find_issues(pred_probs=pred_probs, features=features)
captured = capsys.readouterr()
assert (
"Finding dark, light, low_information, odd_aspect_ratio, odd_size, grayscale, blurry images"
in captured.out
)
assert (
f"from potential issues in the dataset as it exceeds max_prevalence={max_prevalence}"
in captured.out
)
issue_summary = datalab.get_issue_summary()
assert (
len(issue_summary) == 1 + num_datalab_issues
) # adding 1 as no low_information issues present
def test_imagelab_issues_not_checked(
self, image_dataset, pred_probs, features, capsys, num_datalab_issues
):
datalab = Datalab(data=image_dataset, label_name=LABEL_NAME)
datalab.find_issues(pred_probs=pred_probs, features=features)
captured = capsys.readouterr()
assert (
"Finding dark, light, low_information, odd_aspect_ratio, odd_size, grayscale, blurry images"
not in captured.out
)
assert len(datalab.issues) == len(image_dataset)
assert len(datalab.issues.columns) == num_datalab_issues * 2
assert len(datalab.issue_summary) == num_datalab_issues
all_keys = [
"statistics",
"label",
"outlier",
"near_duplicate",
"class_imbalance",
"null",
"underperforming_group",
]
assert set(all_keys) == set(datalab.info.keys())
datalab.report(show_all_issues=True)
captured = capsys.readouterr()
for issue_type in IMAGELAB_ISSUE_TYPES:
assert issue_type not in captured.out
@pytest.mark.usefixtures("set_plt_show")
def test_incremental_issue_check(self, image_dataset, pred_probs, features, capsys):
datalab = Datalab(data=image_dataset, label_name=LABEL_NAME, image_key=IMAGE_NAME)
datalab.find_issues(pred_probs=pred_probs, features=features, issue_types={"label": {}})
assert len(datalab.issues) == len(image_dataset)
assert len(datalab.issues.columns) == 2
assert len(datalab.issue_summary) == 1
all_keys = ["statistics", "label"]
assert set(all_keys) == set(datalab.info.keys())
datalab.report(show_all_issues=True)
captured = capsys.readouterr()
assert "label" in captured.out
datalab.find_issues(issue_types={"image_issue_types": {"dark": {}}})
assert len(datalab.issues) == len(image_dataset)
assert len(datalab.issues.columns) == 4
assert len(datalab.issue_summary) == 2
all_keys = ["statistics", "label", "dark"]
assert set(all_keys) == set(datalab.info.keys())
datalab.report(show_all_issues=True)
captured = capsys.readouterr()
assert "label" in captured.out
assert "dark" in captured.out
with pytest.warns() as record:
datalab.find_issues(
issue_types={"image_issue_types": {"dark": {"threshold": 0.5}, "light": {}}}
)
assert len(record) == 3
assert (
"Overwriting columns ['is_dark_issue', 'dark_score'] in self.issues with columns from imagelab."
== record[0].message.args[0]
)
assert (
"Overwriting ['dark'] rows in self.issue_summary from imagelab."
== record[1].message.args[0]
)
assert "Overwriting key dark in self.info" == record[2].message.args[0]
assert len(datalab.issues) == len(image_dataset)
assert len(datalab.issues.columns) == 6
assert len(datalab.issue_summary) == 3
all_keys = ["statistics", "label", "dark", "light"]
assert set(all_keys) == set(datalab.info.keys())
datalab.report(show_all_issues=True)
captured = capsys.readouterr()
assert "label" in captured.out
assert "dark" in captured.out
@pytest.mark.usefixtures("set_plt_show")
def test_labels_not_required_for_imagelab_issues(
self, image_dataset, features, capsys, num_imagelab_issues
):
datalab = Datalab(data=image_dataset, image_key=IMAGE_NAME)
datalab.find_issues()
captured = capsys.readouterr()
assert (
"Finding dark, light, low_information, odd_aspect_ratio, odd_size, grayscale, blurry images"
in captured.out
)
assert len(datalab.issues) == len(image_dataset)
assert len(datalab.issues.columns) == num_imagelab_issues * 2
assert len(datalab.issue_summary) == num_imagelab_issues
all_keys = IMAGELAB_ISSUE_TYPES + ["statistics"]
assert set(all_keys) == set(datalab.info.keys())
datalab.report(show_all_issues=True)
captured = capsys.readouterr()
for issue_type in IMAGELAB_ISSUE_TYPES:
assert issue_type in captured.out
@pytest.fixture
def lab(self, image_dataset):
lab = Datalab(data=image_dataset, label_name=LABEL_NAME, image_key=IMAGE_NAME)
lab.find_issues()
return lab
def test_get_summary(self, lab):
summary = lab.get_issue_summary("dark")
assert len(summary) == 1
num_issues = summary["num_issues"].values[0]
assert num_issues == 1
@pytest.mark.parametrize(
"list_method", ["list_possible_issue_types", "list_default_issue_types"]
)
def test_list_issue_type_method(self, image_dataset, lab, list_method):
method = getattr(lab, list_method)
issue_types = method()
# Check that Datalab without Imagelab injected has just a subset of possible/default issue types
minimal_lab = Datalab(data=image_dataset)
minimal_method = getattr(minimal_lab, list_method)
datalab_issue_types = minimal_method()
assert set(datalab_issue_types).issubset(set(issue_types))
# The additional issue types found by method should be the same as IMAGELAB_ISSUE_TYPES
assert set(issue_types).difference(datalab_issue_types) == set(IMAGELAB_ISSUE_TYPES)
@pytest.mark.issue1027
def test_get_issues(self, lab):
"""
Test the `get_issues` method of the `lab` object.
This method checks if the columns returned by the `get_issues` method
match the expected columns for each issue type defined in `IMAGELAB_ISSUE_TYPES`.
Raises:
AssertionError: If the columns returned by `get_issues` do not match the expected columns.
"""
test_condition = lambda s: set(lab.get_issues(s).columns) == set(
[f"{s}_score", f"is_{s}_issue"]
)
failed_assertions = [
issue_type for issue_type in IMAGELAB_ISSUE_TYPES if not test_condition(issue_type)
]
assert (
len(failed_assertions) == 0
), f"Tests for `get_issues` with these `issue_types` failed: {failed_assertions}"