test_utils.py 15.2 KB
Newer Older
1
# @author lucasmiranda42
2
3
4
5
6
7
8
9
# encoding: utf-8
# module deepof

"""

Testing module for deepof.utils

"""
10
11

from hypothesis import given
12
from hypothesis import HealthCheck
lucas_miranda's avatar
lucas_miranda committed
13
from hypothesis import settings
14
15
16
17
from hypothesis import strategies as st
from hypothesis.extra.numpy import arrays
from hypothesis.extra.pandas import range_indexes, columns, data_frames
from scipy.spatial import distance
18
from deepof.data import *
lucas_miranda's avatar
lucas_miranda committed
19
from deepof.utils import *
20

21

lucas_miranda's avatar
lucas_miranda committed
22
23
24
25
# AUXILIARY FUNCTIONS #


def autocorr(x, t=1):
lucas_miranda's avatar
lucas_miranda committed
26
    """Computes autocorrelation of the given array with a lag of t"""
lucas_miranda's avatar
lucas_miranda committed
27
28
    return np.round(np.corrcoef(np.array([x[:-t], x[t:]]))[0, 1], 5)

29

30
31
32
# QUALITY CONTROL AND PREPROCESSING #


33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
@settings(deadline=None)
@given(
    v=st.one_of(
        st.just("yes"),
        st.just("true"),
        st.just("t"),
        st.just("y"),
        st.just("1"),
        st.just("no"),
        st.just("false"),
        st.just("f"),
        st.just("n"),
        st.just("0"),
    )
)
def test_str2bool(v):
    assert type(str2bool(v)) == bool


lucas_miranda's avatar
lucas_miranda committed
52
@settings(deadline=None)
lucas_miranda's avatar
lucas_miranda committed
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
@given(
    mult=st.integers(min_value=1, max_value=10),
    dframe=data_frames(
        index=range_indexes(min_size=1),
        columns=columns(["X", "y", "likelihood"], dtype=float),
        rows=st.tuples(
            st.floats(
                min_value=0, max_value=1000, allow_nan=False, allow_infinity=False
            ),
            st.floats(
                min_value=0, max_value=1000, allow_nan=False, allow_infinity=False
            ),
            st.floats(
                min_value=0.01, max_value=1.0, allow_nan=False, allow_infinity=False
            ),
        ),
    ),
    threshold=st.data(),
)
def test_likelihood_qc(mult, dframe, threshold):
    thresh1 = threshold.draw(st.floats(min_value=0.1, max_value=1.0, allow_nan=False))
    thresh2 = threshold.draw(
        st.floats(min_value=thresh1, max_value=1.0, allow_nan=False)
    )

    dframe = pd.concat([dframe] * mult, axis=0)
    idx = pd.MultiIndex.from_product(
        [list(dframe.columns[: len(dframe.columns) // 3]), ["X", "y", "likelihood"]],
        names=["bodyparts", "coords"],
    )
    dframe.columns = idx

    filt1 = likelihood_qc(dframe, thresh1)
    filt2 = likelihood_qc(dframe, thresh2)

    assert np.sum(filt1) <= dframe.shape[0]
    assert np.sum(filt2) <= dframe.shape[0]
    assert np.sum(filt1) >= np.sum(filt2)


lucas_miranda's avatar
lucas_miranda committed
93
@settings(deadline=None)
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
@given(
    tab=data_frames(
        index=range_indexes(min_size=1),
        columns=columns(["X", "y"], dtype=float),
        rows=st.tuples(
            st.floats(
                min_value=0, max_value=1000, allow_nan=False, allow_infinity=False
            ),
            st.floats(
                min_value=0, max_value=1000, allow_nan=False, allow_infinity=False
            ),
        ),
    )
)
def test_bp2polar(tab):
    polar = bp2polar(tab)
    assert np.allclose(polar["rho"], np.sqrt(tab["X"] ** 2 + tab["y"] ** 2))
    assert np.allclose(polar["phi"], np.arctan2(tab["y"], tab["X"]))


lucas_miranda's avatar
lucas_miranda committed
114
@settings(deadline=None)
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
@given(
    mult=st.integers(min_value=1, max_value=10),
    cartdf=data_frames(
        index=range_indexes(min_size=1),
        columns=columns(["X", "y"], dtype=float),
        rows=st.tuples(
            st.floats(
                min_value=0, max_value=1000, allow_nan=False, allow_infinity=False
            ),
            st.floats(
                min_value=0, max_value=1000, allow_nan=False, allow_infinity=False
            ),
        ),
    ),
)
def test_tab2polar(mult, cartdf):
    cart_df = pd.concat([cartdf] * mult, axis=0)
    idx = pd.MultiIndex.from_product(
        [list(cart_df.columns[: len(cart_df.columns) // 2]), ["X", "y"]],
        names=["bodyparts", "coords"],
    )
    cart_df.columns = idx

    assert cart_df.shape == tab2polar(cart_df).shape


lucas_miranda's avatar
lucas_miranda committed
141
@settings(deadline=None)
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
@given(
    pair_array=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=1, max_value=1000),
            st.integers(min_value=4, max_value=4),
        ),
        elements=st.floats(min_value=-1000, max_value=1000, allow_nan=False),
    ),
    arena_abs=st.integers(min_value=1, max_value=1000),
    arena_rel=st.integers(min_value=1, max_value=1000),
)
def test_compute_dist(pair_array, arena_abs, arena_rel):
    assert np.allclose(
        compute_dist(pair_array, arena_abs, arena_rel),
        pd.DataFrame(distance.cdist(pair_array[:, :2], pair_array[:, 2:]).diagonal())
        * arena_abs
        / arena_rel,
    )


lucas_miranda's avatar
lucas_miranda committed
163
@settings(deadline=None)
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
@given(
    cordarray=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=1, max_value=100),
            st.integers(min_value=2, max_value=5).map(lambda x: 4 * x),
        ),
        elements=st.floats(
            min_value=-1000, max_value=1000, allow_nan=False, allow_infinity=False
        ),
    ),
)
def test_bpart_distance(cordarray):
    cord_df = pd.DataFrame(cordarray)
    idx = pd.MultiIndex.from_product(
        [list(cord_df.columns[: len(cord_df.columns) // 2]), ["X", "y"]],
        names=["bodyparts", "coords"],
    )
    cord_df.columns = idx

    bpart = bpart_distance(cord_df)

    assert bpart.shape[0] == cord_df.shape[0]
    assert bpart.shape[1] == len(list(combinations(range(cord_df.shape[1] // 2), 2)))


lucas_miranda's avatar
lucas_miranda committed
190
@settings(deadline=None)
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
@given(
    abc=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=3, max_value=3),
            st.integers(min_value=5, max_value=100),
            st.integers(min_value=2, max_value=2),
        ),
        elements=st.floats(
            min_value=1, max_value=10, allow_nan=False, allow_infinity=False
        ).map(lambda x: x + np.random.uniform(0, 10)),
    ),
)
def test_angle(abc):
    a, b, c = abc

    angles = []
    for i, j, k in zip(a, b, c):
        ang = np.arccos(
            (np.dot(i - j, k - j) / (np.linalg.norm(i - j) * np.linalg.norm(k - j)))
        )
        angles.append(ang)

    print(angle(a, b, c), np.array(angles))

    assert np.allclose(angle(a, b, c), np.array(angles))


lucas_miranda's avatar
lucas_miranda committed
219
@settings(deadline=None)
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
@given(
    array=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=3, max_value=3),
            st.integers(min_value=5, max_value=100),
            st.integers(min_value=2, max_value=2),
        ),
        elements=st.floats(
            min_value=1, max_value=10, allow_nan=False, allow_infinity=False
        ).map(lambda x: x + np.random.uniform(0, 10)),
    )
)
def test_angle_trio(array):
    assert len(angle_trio(array)) == 3


lucas_miranda's avatar
lucas_miranda committed
237
@settings(deadline=None)
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
@given(
    p=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=2, max_value=100),
            st.integers(min_value=2, max_value=2),
        ),
        elements=st.floats(
            min_value=1, max_value=10, allow_nan=False, allow_infinity=False
        ),
    )
)
def test_rotate(p):
    assert np.allclose(rotate(p, 2 * np.pi), p)
    assert np.allclose(rotate(p, np.pi), -p)
lucas_miranda's avatar
lucas_miranda committed
253
    assert np.allclose(rotate(p, 0), p)
254
255


lucas_miranda's avatar
lucas_miranda committed
256
@settings(deadline=None)
257
258
259
260
261
262
263
264
265
266
267
@given(
    data=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=1, max_value=100),
            st.integers(min_value=3, max_value=100),
            st.integers(min_value=1, max_value=10).map(lambda x: 2 * x),
        ),
        elements=st.floats(
            min_value=1, max_value=10, allow_nan=False, allow_infinity=False
        ),
268
    ),
lucas_miranda's avatar
lucas_miranda committed
269
    mode_idx=st.integers(min_value=0, max_value=2),
270
)
271
def test_align_trajectories(data, mode_idx):
lucas_miranda's avatar
lucas_miranda committed
272
    mode = ["center", "all", "none"][mode_idx]
273
274
275
276
277
278
    aligned = align_trajectories(data, mode)
    assert aligned.shape == data.shape
    if mode == "center":
        assert np.allclose(aligned[:, (data.shape[1] - 1) // 2, 0], 0)
    elif mode == "all":
        assert np.allclose(aligned[:, :, 0], 0)
lucas_miranda's avatar
lucas_miranda committed
279
280
    elif mode == "none":
        assert np.allclose(aligned, data)
281
282


lucas_miranda's avatar
lucas_miranda committed
283
@settings(deadline=None)
284
285
286
@given(a=arrays(dtype=bool, shape=st.tuples(st.integers(min_value=3, max_value=1000))))
def test_smooth_boolean_array(a):
    smooth = smooth_boolean_array(a)
lucas_miranda's avatar
lucas_miranda committed
287
288

    def trans(x):
lucas_miranda's avatar
lucas_miranda committed
289
        """In situ function for computing boolean transitions"""
lucas_miranda's avatar
lucas_miranda committed
290
291
        return sum([i + 1 != i for i in range(x.shape[0] - 1)])

292
293
294
    assert trans(a) >= trans(smooth)


lucas_miranda's avatar
lucas_miranda committed
295
@settings(deadline=None)
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
@given(
    a=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=1000, max_value=10000),
            st.integers(min_value=1, max_value=10).map(lambda x: 2 * x),
        ),
        elements=st.floats(
            min_value=1, max_value=10, allow_nan=False, allow_infinity=False
        ),
    ),
    window=st.data(),
)
def test_rolling_window(a, window):
    window_step = window.draw(st.integers(min_value=1, max_value=10))
    window_size = window.draw(
        st.integers(min_value=1, max_value=10).map(lambda x: x * window_step)
    )

    rolled_shape = rolling_window(a, window_size, window_step).shape

317
    assert len(rolled_shape) == len(a.shape) + 1
318
    assert rolled_shape[1] == window_size
lucas_miranda's avatar
lucas_miranda committed
319
320
321
322
323
324
325


@settings(deadline=None)
@given(
    alpha=st.data(),
    series=arrays(
        dtype=float,
326
327
328
        shape=st.tuples(
            st.integers(min_value=10, max_value=1000),
        ),
lucas_miranda's avatar
lucas_miranda committed
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
        elements=st.floats(
            min_value=1.0, max_value=1.0, allow_nan=False, allow_infinity=False
        ),
    ),
)
def test_smooth_mult_trajectory(alpha, series):
    alpha1 = alpha.draw(
        st.floats(min_value=0.1, max_value=1.0, allow_nan=False, allow_infinity=False)
    )
    alpha2 = alpha.draw(
        st.floats(
            min_value=alpha1, max_value=1.0, allow_nan=False, allow_infinity=False
        )
    )

    series *= +np.random.normal(0, 1, len(series))

    smoothed1 = smooth_mult_trajectory(series, alpha1)
    smoothed2 = smooth_mult_trajectory(series, alpha2)

    assert autocorr(smoothed1) >= autocorr(series)
    assert autocorr(smoothed2) >= autocorr(series)
    assert autocorr(smoothed2) <= autocorr(smoothed1)
352
353


354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
@given(mode=st.one_of(st.just("and"), st.just("or")))
def test_interpolate_outliers(mode):

    prun = project(
        path=os.path.join(".", "tests", "test_examples", "test_single_topview"),
        arena="circular",
        arena_dims=tuple([380]),
        video_format=".mp4",
        table_format=".h5",
        exp_conditions={"test": "test_cond"},
    ).run()
    coords = prun.get_coords()
    lkhood = prun.get_quality()
    coords_name = list(coords.keys())[0]

    interp = interpolate_outliers(
        coords[coords_name], lkhood[coords_name], 0.9, exclude="Center", mode=mode
    )
    assert (
        full_outlier_mask(
            interp,
            lkhood,
            likelihood_tolerance=0.9,
            exclude="Center",
            lag=5,
            n_std=2,
            mode=mode,
        ).sum()
        < full_outlier_mask(
            coords,
            lkhood,
            likelihood_tolerance=0.9,
            exclude="Center",
            lag=5,
            n_std=2,
            mode=mode,
        ).sum()
    )
392
393


394
@settings(deadline=None)
395
396
@given(indexes=st.data())
def test_recognize_arena_and_subfunctions(indexes):
lucas_miranda's avatar
lucas_miranda committed
397

398
    path = os.path.join(".", "tests", "test_examples", "test_single_topview", "Videos")
lucas_miranda's avatar
lucas_miranda committed
399
400
401
402
403
    videos = [i for i in os.listdir(path) if i.endswith("mp4")]

    vid_index = indexes.draw(st.integers(min_value=0, max_value=len(videos) - 1))
    recoglimit = indexes.draw(st.integers(min_value=1, max_value=10))

404
    assert recognize_arena(videos, vid_index, path, recoglimit, "")[0] == 0
405
    assert len(recognize_arena(videos, vid_index, path, recoglimit, "circular")) == 3
406
407
408
409
410
411
412
    assert len(recognize_arena(videos, vid_index, path, recoglimit, "circular")[0]) == 3
    assert (
        type(recognize_arena(videos, vid_index, path, recoglimit, "circular")[1]) == int
    )
    assert (
        type(recognize_arena(videos, vid_index, path, recoglimit, "circular")[2]) == int
    )
413
414


lucas_miranda's avatar
lucas_miranda committed
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
@settings(deadline=None)
@given(
    dframe=data_frames(
        index=range_indexes(min_size=50),
        columns=columns(["X1", "y1", "X2", "y2"], dtype=float),
        rows=st.tuples(
            st.floats(min_value=1, max_value=10),
            st.floats(min_value=1, max_value=10),
            st.floats(min_value=1, max_value=10),
            st.floats(min_value=1, max_value=10),
        ),
    ),
    sampler=st.data(),
)
def test_rolling_speed(dframe, sampler):

    dframe *= np.random.uniform(0, 1, dframe.shape)

    order1 = sampler.draw(st.integers(min_value=1, max_value=3))
    order2 = sampler.draw(st.integers(min_value=order1, max_value=3))

    idx = pd.MultiIndex.from_product(
437
438
        [["bpart1", "bpart2"], ["X", "y"]],
        names=["bodyparts", "coords"],
lucas_miranda's avatar
lucas_miranda committed
439
440
441
442
443
444
445
446
447
    )
    dframe.columns = idx

    speeds1 = rolling_speed(dframe, 5, 10, order1)
    speeds2 = rolling_speed(dframe, 5, 10, order2)

    assert speeds1.shape[0] == dframe.shape[0]
    assert speeds1.shape[1] == dframe.shape[1] // 2
    assert np.all(np.std(speeds1) >= np.std(speeds2))
lucas_miranda's avatar
lucas_miranda committed
448
449


450
@settings(
451
452
    deadline=None,
    suppress_health_check=[HealthCheck.too_slow],
453
454
455
456
457
458
459
460
)
@given(
    x=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=10, max_value=1000),
            st.integers(min_value=10, max_value=1000),
        ),
461
462
463
464
        elements=st.floats(
            min_value=1.0,
            max_value=1.0,
        ),
465
466
467
468
469
470
471
472
473
474
    ).map(lambda x: x * np.random.uniform(0, 2, x.shape)),
    n_components=st.integers(min_value=1, max_value=10),
    cv_type=st.integers(min_value=0, max_value=3),
)
def test_gmm_compute(x, n_components, cv_type):
    cv_type = ["spherical", "tied", "diag", "full"][cv_type]
    assert len(gmm_compute(x, n_components, cv_type)) == 2


@settings(
475
476
    deadline=None,
    suppress_health_check=[HealthCheck.too_slow],
477
478
479
480
481
482
483
484
)
@given(
    x=arrays(
        dtype=float,
        shape=st.tuples(
            st.integers(min_value=10, max_value=1000),
            st.integers(min_value=10, max_value=1000),
        ),
485
486
487
488
        elements=st.floats(
            min_value=1.0,
            max_value=1.0,
        ),
489
490
491
492
493
494
495
496
497
498
499
500
501
502
    ).map(lambda x: x * np.random.uniform(0, 2, x.shape)),
    sampler=st.data(),
)
def test_gmm_model_selection(x, sampler):
    n_component_range = range(1, sampler.draw(st.integers(min_value=2, max_value=5)))
    part_size = sampler.draw(
        st.integers(min_value=x.shape[0] // 2, max_value=x.shape[0] * 2)
    )
    assert (
        len(
            gmm_model_selection(pd.DataFrame(x), n_component_range, part_size, n_runs=1)
        )
        == 3
    )
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536


@settings(deadline=None)
@given(sampler=st.data(), autocorrelation=st.booleans(), return_graph=st.booleans())
def test_cluster_transition_matrix(sampler, autocorrelation, return_graph):

    nclusts = sampler.draw(st.integers(min_value=1, max_value=10))
    cluster_sequence = sampler.draw(
        arrays(
            dtype=int,
            shape=st.tuples(st.integers(min_value=10, max_value=1000)),
            elements=st.integers(min_value=1, max_value=nclusts),
        ).filter(lambda x: len(set(x)) != 1)
    )

    trans = cluster_transition_matrix(
        cluster_sequence, nclusts, autocorrelation, return_graph
    )

    if autocorrelation:
        assert len(trans) == 2

        if return_graph:
            assert type(trans[0]) == nx.Graph
        else:
            assert type(trans[0]) == np.ndarray

        assert type(trans[1]) == np.ndarray

    else:
        if return_graph:
            assert type(trans) == nx.Graph
        else:
            assert type(trans) == np.ndarray