test_transforms.py 85.7 KB
Newer Older
1
import itertools
2
import os
3
4
import torch
import torchvision.transforms as transforms
5
import torchvision.transforms.functional as F
6
import torchvision.transforms.functional_tensor as F_t
7
from torch._utils_internal import get_file_path_2
8
from numpy.testing import assert_array_almost_equal
9
import unittest
10
import math
11
import random
12
import numpy as np
13
14
15
16
17
18
from PIL import Image
try:
    import accimage
except ImportError:
    accimage = None

19
20
21
22
23
try:
    from scipy import stats
except ImportError:
    stats = None

24
from common_utils import cycle_over, int_dtypes, float_dtypes
25
26


27
GRACE_HOPPER = get_file_path_2(
28
    os.path.dirname(os.path.abspath(__file__)), 'assets', 'encode_jpeg', 'grace_hopper_517x606.jpg')
29
30


31
class Tester(unittest.TestCase):
32

33
    def test_center_crop(self):
34
35
36
        height = random.randint(10, 32) * 2
        width = random.randint(10, 32) * 2
        oheight = random.randint(5, (height - 2) / 2) * 2
37
38
        owidth = random.randint(5, (width - 2) / 2) * 2

39
        img = torch.ones(3, height, width)
40
41
42
        oh1 = (height - oheight) // 2
        ow1 = (width - owidth) // 2
        imgnarrow = img[:, oh1:oh1 + oheight, ow1:ow1 + owidth]
43
44
45
46
47
48
        imgnarrow.fill_(0)
        result = transforms.Compose([
            transforms.ToPILImage(),
            transforms.CenterCrop((oheight, owidth)),
            transforms.ToTensor(),
        ])(img)
49
50
        self.assertEqual(result.sum(), 0,
                         "height: {} width: {} oheight: {} owdith: {}".format(height, width, oheight, owidth))
51
52
53
54
55
56
57
58
        oheight += 1
        owidth += 1
        result = transforms.Compose([
            transforms.ToPILImage(),
            transforms.CenterCrop((oheight, owidth)),
            transforms.ToTensor(),
        ])(img)
        sum1 = result.sum()
59
60
        self.assertGreater(sum1, 1,
                           "height: {} width: {} oheight: {} owdith: {}".format(height, width, oheight, owidth))
61
        oheight += 1
62
        owidth += 1
63
64
65
66
67
68
        result = transforms.Compose([
            transforms.ToPILImage(),
            transforms.CenterCrop((oheight, owidth)),
            transforms.ToTensor(),
        ])(img)
        sum2 = result.sum()
69
70
71
72
        self.assertGreater(sum2, 0,
                           "height: {} width: {} oheight: {} owdith: {}".format(height, width, oheight, owidth))
        self.assertGreater(sum2, sum1,
                           "height: {} width: {} oheight: {} owdith: {}".format(height, width, oheight, owidth))
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
    def test_center_crop_2(self):
        """ Tests when center crop size is larger than image size, along any dimension"""
        even_image_size = (random.randint(10, 32) * 2, random.randint(10, 32) * 2)
        odd_image_size = (even_image_size[0] + 1, even_image_size[1] + 1)

        # Since height is independent of width, we can ignore images with odd height and even width and vice-versa.
        input_image_sizes = [even_image_size, odd_image_size]

        # Get different crop sizes
        delta = random.choice((1, 3, 5))
        crop_size_delta = [-2 * delta, -delta, 0, delta, 2 * delta]
        crop_size_params = itertools.product(input_image_sizes, crop_size_delta, crop_size_delta)

        for (input_image_size, delta_height, delta_width) in crop_size_params:
            img = torch.ones(3, *input_image_size)
            crop_size = (input_image_size[0] + delta_height, input_image_size[1] + delta_width)

            # Test both transforms, one with PIL input and one with tensor
            output_pil = transforms.Compose([
                transforms.ToPILImage(),
                transforms.CenterCrop(crop_size),
                transforms.ToTensor()],
            )(img)
            self.assertEqual(output_pil.size()[1:3], crop_size,
                             "image_size: {} crop_size: {}".format(input_image_size, crop_size))

            output_tensor = transforms.CenterCrop(crop_size)(img)
            self.assertEqual(output_tensor.size()[1:3], crop_size,
                             "image_size: {} crop_size: {}".format(input_image_size, crop_size))

            # Ensure output for PIL and Tensor are equal
            self.assertEqual((output_tensor - output_pil).sum(), 0,
                             "image_size: {} crop_size: {}".format(input_image_size, crop_size))

            # Check if content in center of both image and cropped output is same.
            center_size = (min(crop_size[0], input_image_size[0]), min(crop_size[1], input_image_size[1]))
            crop_center_tl, input_center_tl = [0, 0], [0, 0]
            for index in range(2):
                if crop_size[index] > input_image_size[index]:
                    crop_center_tl[index] = (crop_size[index] - input_image_size[index]) // 2
                else:
                    input_center_tl[index] = (input_image_size[index] - crop_size[index]) // 2

            output_center = output_pil[
                :,
                crop_center_tl[0]:crop_center_tl[0] + center_size[0],
                crop_center_tl[1]:crop_center_tl[1] + center_size[1]
            ]

            img_center = img[
                :,
                input_center_tl[0]:input_center_tl[0] + center_size[0],
                input_center_tl[1]:input_center_tl[1] + center_size[1]
            ]

            self.assertEqual((output_center - img_center).sum(), 0,
                             "image_size: {} crop_size: {}".format(input_image_size, crop_size))

132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
    def test_five_crop(self):
        to_pil_image = transforms.ToPILImage()
        h = random.randint(5, 25)
        w = random.randint(5, 25)
        for single_dim in [True, False]:
            crop_h = random.randint(1, h)
            crop_w = random.randint(1, w)
            if single_dim:
                crop_h = min(crop_h, crop_w)
                crop_w = crop_h
                transform = transforms.FiveCrop(crop_h)
            else:
                transform = transforms.FiveCrop((crop_h, crop_w))

            img = torch.FloatTensor(3, h, w).uniform_()
            results = transform(to_pil_image(img))

149
            self.assertEqual(len(results), 5)
150
            for crop in results:
151
                self.assertEqual(crop.size, (crop_w, crop_h))
152
153
154
155
156
157
158
159

            to_pil_image = transforms.ToPILImage()
            tl = to_pil_image(img[:, 0:crop_h, 0:crop_w])
            tr = to_pil_image(img[:, 0:crop_h, w - crop_w:])
            bl = to_pil_image(img[:, h - crop_h:, 0:crop_w])
            br = to_pil_image(img[:, h - crop_h:, w - crop_w:])
            center = transforms.CenterCrop((crop_h, crop_w))(to_pil_image(img))
            expected_output = (tl, tr, bl, br, center)
160
            self.assertEqual(results, expected_output)
161
162
163
164
165
166
167
168
169
170
171
172

    def test_ten_crop(self):
        to_pil_image = transforms.ToPILImage()
        h = random.randint(5, 25)
        w = random.randint(5, 25)
        for should_vflip in [True, False]:
            for single_dim in [True, False]:
                crop_h = random.randint(1, h)
                crop_w = random.randint(1, w)
                if single_dim:
                    crop_h = min(crop_h, crop_w)
                    crop_w = crop_h
173
174
                    transform = transforms.TenCrop(crop_h,
                                                   vertical_flip=should_vflip)
175
176
                    five_crop = transforms.FiveCrop(crop_h)
                else:
177
178
                    transform = transforms.TenCrop((crop_h, crop_w),
                                                   vertical_flip=should_vflip)
179
180
181
182
183
                    five_crop = transforms.FiveCrop((crop_h, crop_w))

                img = to_pil_image(torch.FloatTensor(3, h, w).uniform_())
                results = transform(img)
                expected_output = five_crop(img)
184
185
186
187
188

                # Checking if FiveCrop and TenCrop can be printed as string
                transform.__repr__()
                five_crop.__repr__()

189
190
191
192
193
194
195
                if should_vflip:
                    vflipped_img = img.transpose(Image.FLIP_TOP_BOTTOM)
                    expected_output += five_crop(vflipped_img)
                else:
                    hflipped_img = img.transpose(Image.FLIP_LEFT_RIGHT)
                    expected_output += five_crop(hflipped_img)

196
197
                self.assertEqual(len(results), 10)
                self.assertEqual(results, expected_output)
198

199
200
201
202
203
204
205
206
    def test_randomresized_params(self):
        height = random.randint(24, 32) * 2
        width = random.randint(24, 32) * 2
        img = torch.ones(3, height, width)
        to_pil_image = transforms.ToPILImage()
        img = to_pil_image(img)
        size = 100
        epsilon = 0.05
207
        min_scale = 0.25
Francisco Massa's avatar
Francisco Massa committed
208
        for _ in range(10):
209
            scale_min = max(round(random.random(), 2), min_scale)
210
            scale_range = (scale_min, scale_min + round(random.random(), 2))
211
            aspect_min = max(round(random.random(), 2), epsilon)
212
213
            aspect_ratio_range = (aspect_min, aspect_min + round(random.random(), 2))
            randresizecrop = transforms.RandomResizedCrop(size, scale_range, aspect_ratio_range)
214
            i, j, h, w = randresizecrop.get_params(img, scale_range, aspect_ratio_range)
215
            aspect_ratio_obtained = w / h
216
217
218
219
220
221
222
            self.assertTrue((min(aspect_ratio_range) - epsilon <= aspect_ratio_obtained and
                             aspect_ratio_obtained <= max(aspect_ratio_range) + epsilon) or
                            aspect_ratio_obtained == 1.0)
            self.assertIsInstance(i, int)
            self.assertIsInstance(j, int)
            self.assertIsInstance(h, int)
            self.assertIsInstance(w, int)
223

224
    def test_randomperspective(self):
Francisco Massa's avatar
Francisco Massa committed
225
        for _ in range(10):
226
227
228
229
230
231
232
233
234
235
            height = random.randint(24, 32) * 2
            width = random.randint(24, 32) * 2
            img = torch.ones(3, height, width)
            to_pil_image = transforms.ToPILImage()
            img = to_pil_image(img)
            perp = transforms.RandomPerspective()
            startpoints, endpoints = perp.get_params(width, height, 0.5)
            tr_img = F.perspective(img, startpoints, endpoints)
            tr_img2 = F.to_tensor(F.perspective(tr_img, endpoints, startpoints))
            tr_img = F.to_tensor(tr_img)
236
237
238
239
            self.assertEqual(img.size[0], width)
            self.assertEqual(img.size[1], height)
            self.assertGreater(torch.nn.functional.mse_loss(tr_img, F.to_tensor(img)) + 0.3,
                               torch.nn.functional.mse_loss(tr_img2, F.to_tensor(img)))
240

241
    def test_randomperspective_fill(self):
242
243
244
245
246
247
248
249

        # assert fill being either a Sequence or a Number
        with self.assertRaises(TypeError):
            transforms.RandomPerspective(fill={})

        t = transforms.RandomPerspective(fill=None)
        self.assertTrue(t.fill == 0)

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
        height = 100
        width = 100
        img = torch.ones(3, height, width)
        to_pil_image = transforms.ToPILImage()
        img = to_pil_image(img)

        modes = ("L", "RGB", "F")
        nums_bands = [len(mode) for mode in modes]
        fill = 127

        for mode, num_bands in zip(modes, nums_bands):
            img_conv = img.convert(mode)
            perspective = transforms.RandomPerspective(p=1, fill=fill)
            tr_img = perspective(img_conv)
            pixel = tr_img.getpixel((0, 0))

            if not isinstance(pixel, tuple):
                pixel = (pixel,)
            self.assertTupleEqual(pixel, tuple([fill] * num_bands))

        for mode, num_bands in zip(modes, nums_bands):
            img_conv = img.convert(mode)
            startpoints, endpoints = transforms.RandomPerspective.get_params(width, height, 0.5)
            tr_img = F.perspective(img_conv, startpoints, endpoints, fill=fill)
            pixel = tr_img.getpixel((0, 0))
275

276
277
278
279
280
281
282
283
            if not isinstance(pixel, tuple):
                pixel = (pixel,)
            self.assertTupleEqual(pixel, tuple([fill] * num_bands))

            for wrong_num_bands in set(nums_bands) - {num_bands}:
                with self.assertRaises(ValueError):
                    F.perspective(img_conv, startpoints, endpoints, fill=tuple([fill] * wrong_num_bands))

284
    def test_resize(self):
285

286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
        input_sizes = [
            # height, width
            # square image
            (28, 28),
            (27, 27),
            # rectangular image: h < w
            (28, 34),
            (29, 35),
            # rectangular image: h > w
            (34, 28),
            (35, 29),
        ]
        test_output_sizes_1 = [
            # single integer
            22, 27, 28, 36,
            # single integer in tuple/list
            [22, ], (27, ),
        ]
        test_output_sizes_2 = [
            # two integers
            [22, 22], [22, 28], [22, 36],
            [27, 22], [36, 22], [28, 28],
            [28, 37], [37, 27], [37, 37]
        ]

        for height, width in input_sizes:
            img = Image.new("RGB", size=(width, height), color=127)

            for osize in test_output_sizes_1:
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
                for max_size in (None, 37, 1000):

                    t = transforms.Resize(osize, max_size=max_size)
                    result = t(img)

                    msg = "{}, {} - {} - {}".format(height, width, osize, max_size)
                    osize = osize[0] if isinstance(osize, (list, tuple)) else osize
                    # If size is an int, smaller edge of the image will be matched to this number.
                    # i.e, if height > width, then image will be rescaled to (size * height / width, size).
                    if height < width:
                        exp_w, exp_h = (int(osize * width / height), osize)  # (w, h)
                        if max_size is not None and max_size < exp_w:
                            exp_w, exp_h = max_size, int(max_size * exp_h / exp_w)
                        self.assertEqual(result.size, (exp_w, exp_h), msg=msg)
                    elif width < height:
                        exp_w, exp_h = (osize, int(osize * height / width))  # (w, h)
                        if max_size is not None and max_size < exp_h:
                            exp_w, exp_h = int(max_size * exp_w / exp_h), max_size
                        self.assertEqual(result.size, (exp_w, exp_h), msg=msg)
                    else:
                        exp_w, exp_h = (osize, osize)  # (w, h)
                        if max_size is not None and max_size < osize:
                            exp_w, exp_h = max_size, max_size
                        self.assertEqual(result.size, (exp_w, exp_h), msg=msg)
339

340
341
        for height, width in input_sizes:
            img = Image.new("RGB", size=(width, height), color=127)
342

343
344
            for osize in test_output_sizes_2:
                oheight, owidth = osize
345

346
347
                t = transforms.Resize(osize)
                result = t(img)
348

349
                self.assertEqual((owidth, oheight), result.size)
350

351
352
353
354
    def test_random_crop(self):
        height = random.randint(10, 32) * 2
        width = random.randint(10, 32) * 2
        oheight = random.randint(5, (height - 2) / 2) * 2
355
        owidth = random.randint(5, (width - 2) / 2) * 2
356
357
358
359
360
361
        img = torch.ones(3, height, width)
        result = transforms.Compose([
            transforms.ToPILImage(),
            transforms.RandomCrop((oheight, owidth)),
            transforms.ToTensor(),
        ])(img)
362
363
        self.assertEqual(result.size(1), oheight)
        self.assertEqual(result.size(2), owidth)
364

365
366
367
368
369
370
        padding = random.randint(1, 20)
        result = transforms.Compose([
            transforms.ToPILImage(),
            transforms.RandomCrop((oheight, owidth), padding=padding),
            transforms.ToTensor(),
        ])(img)
371
372
        self.assertEqual(result.size(1), oheight)
        self.assertEqual(result.size(2), owidth)
373

374
375
376
377
378
        result = transforms.Compose([
            transforms.ToPILImage(),
            transforms.RandomCrop((height, width)),
            transforms.ToTensor()
        ])(img)
379
380
381
        self.assertEqual(result.size(1), height)
        self.assertEqual(result.size(2), width)
        self.assertTrue(np.allclose(img.numpy(), result.numpy()))
382

383
384
385
386
387
        result = transforms.Compose([
            transforms.ToPILImage(),
            transforms.RandomCrop((height + 1, width + 1), pad_if_needed=True),
            transforms.ToTensor(),
        ])(img)
388
389
        self.assertEqual(result.size(1), height + 1)
        self.assertEqual(result.size(2), width + 1)
390

vfdev's avatar
vfdev committed
391
392
393
394
395
        t = transforms.RandomCrop(48)
        img = torch.ones(3, 32, 32)
        with self.assertRaisesRegex(ValueError, r"Required crop size .+ is larger then input image size .+"):
            t(img)

396
397
398
399
400
    def test_pad(self):
        height = random.randint(10, 32) * 2
        width = random.randint(10, 32) * 2
        img = torch.ones(3, height, width)
        padding = random.randint(1, 20)
401
        fill = random.randint(1, 50)
402
403
        result = transforms.Compose([
            transforms.ToPILImage(),
404
            transforms.Pad(padding, fill=fill),
405
406
            transforms.ToTensor(),
        ])(img)
407
408
        self.assertEqual(result.size(1), height + 2 * padding)
        self.assertEqual(result.size(2), width + 2 * padding)
409
410
411
412
413
414
415
416
        # check that all elements in the padded region correspond
        # to the pad value
        fill_v = fill / 255
        eps = 1e-5
        self.assertTrue((result[:, :padding, :] - fill_v).abs().max() < eps)
        self.assertTrue((result[:, :, :padding] - fill_v).abs().max() < eps)
        self.assertRaises(ValueError, transforms.Pad(padding, fill=(1, 2)),
                          transforms.ToPILImage()(img))
Soumith Chintala's avatar
Soumith Chintala committed
417

418
419
420
421
422
423
424
    def test_pad_with_tuple_of_pad_values(self):
        height = random.randint(10, 32) * 2
        width = random.randint(10, 32) * 2
        img = transforms.ToPILImage()(torch.ones(3, height, width))

        padding = tuple([random.randint(1, 20) for _ in range(2)])
        output = transforms.Pad(padding)(img)
425
        self.assertEqual(output.size, (width + padding[0] * 2, height + padding[1] * 2))
426
427
428

        padding = tuple([random.randint(1, 20) for _ in range(4)])
        output = transforms.Pad(padding)(img)
429
430
        self.assertEqual(output.size[0], width + padding[0] + padding[2])
        self.assertEqual(output.size[1], height + padding[1] + padding[3])
431

432
433
434
        # Checking if Padding can be printed as string
        transforms.Pad(padding).__repr__()

435
436
    def test_pad_with_non_constant_padding_modes(self):
        """Unit tests for edge, reflect, symmetric padding"""
vfdev's avatar
vfdev committed
437
        img = torch.zeros(3, 27, 27).byte()
438
439
440
441
442
443
444
445
446
        img[:, :, 0] = 1  # Constant value added to leftmost edge
        img = transforms.ToPILImage()(img)
        img = F.pad(img, 1, (200, 200, 200))

        # pad 3 to all sidess
        edge_padded_img = F.pad(img, 3, padding_mode='edge')
        # First 6 elements of leftmost edge in the middle of the image, values are in order:
        # edge_pad, edge_pad, edge_pad, constant_pad, constant value added to leftmost edge, 0
        edge_middle_slice = np.asarray(edge_padded_img).transpose(2, 0, 1)[0][17][:6]
447
448
        self.assertTrue(np.all(edge_middle_slice == np.asarray([200, 200, 200, 200, 1, 0])))
        self.assertEqual(transforms.ToTensor()(edge_padded_img).size(), (3, 35, 35))
449
450
451
452
453
454

        # Pad 3 to left/right, 2 to top/bottom
        reflect_padded_img = F.pad(img, (3, 2), padding_mode='reflect')
        # First 6 elements of leftmost edge in the middle of the image, values are in order:
        # reflect_pad, reflect_pad, reflect_pad, constant_pad, constant value added to leftmost edge, 0
        reflect_middle_slice = np.asarray(reflect_padded_img).transpose(2, 0, 1)[0][17][:6]
455
456
        self.assertTrue(np.all(reflect_middle_slice == np.asarray([0, 0, 1, 200, 1, 0])))
        self.assertEqual(transforms.ToTensor()(reflect_padded_img).size(), (3, 33, 35))
457
458
459
460
461
462

        # Pad 3 to left, 2 to top, 2 to right, 1 to bottom
        symmetric_padded_img = F.pad(img, (3, 2, 2, 1), padding_mode='symmetric')
        # First 6 elements of leftmost edge in the middle of the image, values are in order:
        # sym_pad, sym_pad, sym_pad, constant_pad, constant value added to leftmost edge, 0
        symmetric_middle_slice = np.asarray(symmetric_padded_img).transpose(2, 0, 1)[0][17][:6]
463
464
        self.assertTrue(np.all(symmetric_middle_slice == np.asarray([0, 1, 200, 200, 1, 0])))
        self.assertEqual(transforms.ToTensor()(symmetric_padded_img).size(), (3, 32, 34))
465

466
467
468
469
470
471
472
473
474
475
        # Check negative padding explicitly for symmetric case, since it is not
        # implemented for tensor case to compare to
        # Crop 1 to left, pad 2 to top, pad 3 to right, crop 3 to bottom
        symmetric_padded_img_neg = F.pad(img, (-1, 2, 3, -3), padding_mode='symmetric')
        symmetric_neg_middle_left = np.asarray(symmetric_padded_img_neg).transpose(2, 0, 1)[0][17][:3]
        symmetric_neg_middle_right = np.asarray(symmetric_padded_img_neg).transpose(2, 0, 1)[0][17][-4:]
        self.assertTrue(np.all(symmetric_neg_middle_left == np.asarray([1, 0, 0])))
        self.assertTrue(np.all(symmetric_neg_middle_right == np.asarray([200, 200, 0, 0])))
        self.assertEqual(transforms.ToTensor()(symmetric_padded_img_neg).size(), (3, 28, 31))

476
    def test_pad_raises_with_invalid_pad_sequence_len(self):
477
478
479
480
481
482
483
484
485
        with self.assertRaises(ValueError):
            transforms.Pad(())

        with self.assertRaises(ValueError):
            transforms.Pad((1, 2, 3))

        with self.assertRaises(ValueError):
            transforms.Pad((1, 2, 3, 4, 5))

486
487
488
489
490
491
492
493
    def test_pad_with_mode_F_images(self):
        pad = 2
        transform = transforms.Pad(pad)

        img = Image.new("F", (10, 10))
        padded_img = transform(img)
        self.assertSequenceEqual(padded_img.size, [edge_size + 2 * pad for edge_size in img.size])

Soumith Chintala's avatar
Soumith Chintala committed
494
495
496
497
    def test_lambda(self):
        trans = transforms.Lambda(lambda x: x.add(10))
        x = torch.randn(10)
        y = trans(x)
498
        self.assertTrue(y.equal(torch.add(x, 10)))
Soumith Chintala's avatar
Soumith Chintala committed
499
500
501
502

        trans = transforms.Lambda(lambda x: x.add_(10))
        x = torch.randn(10)
        y = trans(x)
503
        self.assertTrue(y.equal(x))
504

505
506
507
        # Checking if Lambda can be printed as string
        trans.__repr__()

508
    @unittest.skipIf(stats is None, 'scipy.stats not available')
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
    def test_random_apply(self):
        random_state = random.getstate()
        random.seed(42)
        random_apply_transform = transforms.RandomApply(
            [
                transforms.RandomRotation((-45, 45)),
                transforms.RandomHorizontalFlip(),
                transforms.RandomVerticalFlip(),
            ], p=0.75
        )
        img = transforms.ToPILImage()(torch.rand(3, 10, 10))
        num_samples = 250
        num_applies = 0
        for _ in range(num_samples):
            out = random_apply_transform(img)
            if out != img:
                num_applies += 1

        p_value = stats.binom_test(num_applies, num_samples, p=0.75)
        random.setstate(random_state)
529
        self.assertGreater(p_value, 0.0001)
530
531
532
533

        # Checking if RandomApply can be printed as string
        random_apply_transform.__repr__()

534
    @unittest.skipIf(stats is None, 'scipy.stats not available')
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
    def test_random_choice(self):
        random_state = random.getstate()
        random.seed(42)
        random_choice_transform = transforms.RandomChoice(
            [
                transforms.Resize(15),
                transforms.Resize(20),
                transforms.CenterCrop(10)
            ]
        )
        img = transforms.ToPILImage()(torch.rand(3, 25, 25))
        num_samples = 250
        num_resize_15 = 0
        num_resize_20 = 0
        num_crop_10 = 0
        for _ in range(num_samples):
            out = random_choice_transform(img)
            if out.size == (15, 15):
                num_resize_15 += 1
            elif out.size == (20, 20):
                num_resize_20 += 1
            elif out.size == (10, 10):
                num_crop_10 += 1

        p_value = stats.binom_test(num_resize_15, num_samples, p=0.33333)
560
        self.assertGreater(p_value, 0.0001)
561
        p_value = stats.binom_test(num_resize_20, num_samples, p=0.33333)
562
        self.assertGreater(p_value, 0.0001)
563
        p_value = stats.binom_test(num_crop_10, num_samples, p=0.33333)
564
        self.assertGreater(p_value, 0.0001)
565
566
567
568
569

        random.setstate(random_state)
        # Checking if RandomChoice can be printed as string
        random_choice_transform.__repr__()

570
    @unittest.skipIf(stats is None, 'scipy.stats not available')
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
    def test_random_order(self):
        random_state = random.getstate()
        random.seed(42)
        random_order_transform = transforms.RandomOrder(
            [
                transforms.Resize(20),
                transforms.CenterCrop(10)
            ]
        )
        img = transforms.ToPILImage()(torch.rand(3, 25, 25))
        num_samples = 250
        num_normal_order = 0
        resize_crop_out = transforms.CenterCrop(10)(transforms.Resize(20)(img))
        for _ in range(num_samples):
            out = random_order_transform(img)
            if out == resize_crop_out:
                num_normal_order += 1

        p_value = stats.binom_test(num_normal_order, num_samples, p=0.5)
        random.setstate(random_state)
591
        self.assertGreater(p_value, 0.0001)
592
593
594
595

        # Checking if RandomOrder can be printed as string
        random_order_transform.__repr__()

596
    def test_to_tensor(self):
597
        test_channels = [1, 3, 4]
598
599
        height, width = 4, 4
        trans = transforms.ToTensor()
600

601
602
603
604
605
606
607
        with self.assertRaises(TypeError):
            trans(np.random.rand(1, height, width).tolist())

        with self.assertRaises(ValueError):
            trans(np.random.rand(height))
            trans(np.random.rand(1, 1, height, width))

608
609
610
611
        for channels in test_channels:
            input_data = torch.ByteTensor(channels, height, width).random_(0, 255).float().div_(255)
            img = transforms.ToPILImage()(input_data)
            output = trans(img)
612
            self.assertTrue(np.allclose(input_data.numpy(), output.numpy()))
613

614
            ndarray = np.random.randint(low=0, high=255, size=(height, width, channels)).astype(np.uint8)
615
616
            output = trans(ndarray)
            expected_output = ndarray.transpose((2, 0, 1)) / 255.0
617
            self.assertTrue(np.allclose(output.numpy(), expected_output))
618

619
620
621
            ndarray = np.random.rand(height, width, channels).astype(np.float32)
            output = trans(ndarray)
            expected_output = ndarray.transpose((2, 0, 1))
622
            self.assertTrue(np.allclose(output.numpy(), expected_output))
623

624
625
626
627
        # separate test for mode '1' PIL images
        input_data = torch.ByteTensor(1, height, width).bernoulli_()
        img = transforms.ToPILImage()(input_data.mul(255)).convert('1')
        output = trans(img)
628
        self.assertTrue(np.allclose(input_data.numpy(), output.numpy()))
629

630
631
632
633
634
635
636
637
638
639
640
641
642
643
    def test_to_tensor_with_other_default_dtypes(self):
        current_def_dtype = torch.get_default_dtype()

        t = transforms.ToTensor()
        np_arr = np.random.randint(0, 255, (32, 32, 3), dtype=np.uint8)
        img = Image.fromarray(np_arr)

        for dtype in [torch.float16, torch.float, torch.double]:
            torch.set_default_dtype(dtype)
            res = t(img)
            self.assertTrue(res.dtype == dtype, msg=f"{res.dtype} vs {dtype}")

        torch.set_default_dtype(current_def_dtype)

644
645
646
647
    def test_max_value(self):
        for dtype in int_dtypes():
            self.assertEqual(F_t._max_value(dtype), torch.iinfo(dtype).max)

648
649
650
651
        # remove float testing as it can lead to errors such as
        # runtime error: 5.7896e+76 is outside the range of representable values of type 'float'
        # for dtype in float_dtypes():
        #     self.assertGreater(F_t._max_value(dtype), torch.finfo(dtype).max)
652

653
654
655
656
657
658
    def test_convert_image_dtype_float_to_float(self):
        for input_dtype, output_dtypes in cycle_over(float_dtypes()):
            input_image = torch.tensor((0.0, 1.0), dtype=input_dtype)
            for output_dtype in output_dtypes:
                with self.subTest(input_dtype=input_dtype, output_dtype=output_dtype):
                    transform = transforms.ConvertImageDtype(output_dtype)
659
660
                    transform_script = torch.jit.script(F.convert_image_dtype)

661
                    output_image = transform(input_image)
662
663
664
665
                    output_image_script = transform_script(input_image, output_dtype)

                    script_diff = output_image_script - output_image
                    self.assertLess(script_diff.abs().max(), 1e-6)
666
667
668
669
670
671
672
673
674
675
676
677
678

                    actual_min, actual_max = output_image.tolist()
                    desired_min, desired_max = 0.0, 1.0

                    self.assertAlmostEqual(actual_min, desired_min)
                    self.assertAlmostEqual(actual_max, desired_max)

    def test_convert_image_dtype_float_to_int(self):
        for input_dtype in float_dtypes():
            input_image = torch.tensor((0.0, 1.0), dtype=input_dtype)
            for output_dtype in int_dtypes():
                with self.subTest(input_dtype=input_dtype, output_dtype=output_dtype):
                    transform = transforms.ConvertImageDtype(output_dtype)
679
                    transform_script = torch.jit.script(F.convert_image_dtype)
680
681
682
683
684
685
686
687

                    if (input_dtype == torch.float32 and output_dtype in (torch.int32, torch.int64)) or (
                            input_dtype == torch.float64 and output_dtype == torch.int64
                    ):
                        with self.assertRaises(RuntimeError):
                            transform(input_image)
                    else:
                        output_image = transform(input_image)
688
689
690
691
                        output_image_script = transform_script(input_image, output_dtype)

                        script_diff = output_image_script - output_image
                        self.assertLess(script_diff.abs().max(), 1e-6)
692
693
694
695
696
697
698
699
700
701
702
703
704

                        actual_min, actual_max = output_image.tolist()
                        desired_min, desired_max = 0, torch.iinfo(output_dtype).max

                        self.assertEqual(actual_min, desired_min)
                        self.assertEqual(actual_max, desired_max)

    def test_convert_image_dtype_int_to_float(self):
        for input_dtype in int_dtypes():
            input_image = torch.tensor((0, torch.iinfo(input_dtype).max), dtype=input_dtype)
            for output_dtype in float_dtypes():
                with self.subTest(input_dtype=input_dtype, output_dtype=output_dtype):
                    transform = transforms.ConvertImageDtype(output_dtype)
705
706
                    transform_script = torch.jit.script(F.convert_image_dtype)

707
                    output_image = transform(input_image)
708
709
710
711
                    output_image_script = transform_script(input_image, output_dtype)

                    script_diff = output_image_script - output_image
                    self.assertLess(script_diff.abs().max(), 1e-6)
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729

                    actual_min, actual_max = output_image.tolist()
                    desired_min, desired_max = 0.0, 1.0

                    self.assertAlmostEqual(actual_min, desired_min)
                    self.assertGreaterEqual(actual_min, desired_min)
                    self.assertAlmostEqual(actual_max, desired_max)
                    self.assertLessEqual(actual_max, desired_max)

    def test_convert_image_dtype_int_to_int(self):
        for input_dtype, output_dtypes in cycle_over(int_dtypes()):
            input_max = torch.iinfo(input_dtype).max
            input_image = torch.tensor((0, input_max), dtype=input_dtype)
            for output_dtype in output_dtypes:
                output_max = torch.iinfo(output_dtype).max

                with self.subTest(input_dtype=input_dtype, output_dtype=output_dtype):
                    transform = transforms.ConvertImageDtype(output_dtype)
730
731
                    transform_script = torch.jit.script(F.convert_image_dtype)

732
                    output_image = transform(input_image)
733
734
735
736
737
738
                    output_image_script = transform_script(input_image, output_dtype)

                    script_diff = output_image_script.float() - output_image.float()
                    self.assertLess(
                        script_diff.abs().max(), 1e-6, msg="{} vs {}".format(output_image_script, output_image)
                    )
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771

                    actual_min, actual_max = output_image.tolist()
                    desired_min, desired_max = 0, output_max

                    # see https://github.com/pytorch/vision/pull/2078#issuecomment-641036236 for details
                    if input_max >= output_max:
                        error_term = 0
                    else:
                        error_term = 1 - (torch.iinfo(output_dtype).max + 1) // (torch.iinfo(input_dtype).max + 1)

                    self.assertEqual(actual_min, desired_min)
                    self.assertEqual(actual_max, desired_max + error_term)

    def test_convert_image_dtype_int_to_int_consistency(self):
        for input_dtype, output_dtypes in cycle_over(int_dtypes()):
            input_max = torch.iinfo(input_dtype).max
            input_image = torch.tensor((0, input_max), dtype=input_dtype)
            for output_dtype in output_dtypes:
                output_max = torch.iinfo(output_dtype).max
                if output_max <= input_max:
                    continue

                with self.subTest(input_dtype=input_dtype, output_dtype=output_dtype):
                    transform = transforms.ConvertImageDtype(output_dtype)
                    inverse_transfrom = transforms.ConvertImageDtype(input_dtype)
                    output_image = inverse_transfrom(transform(input_image))

                    actual_min, actual_max = output_image.tolist()
                    desired_min, desired_max = 0, input_max

                    self.assertEqual(actual_min, desired_min)
                    self.assertEqual(actual_max, desired_max)

772
773
774
775
776
777
778
779
    @unittest.skipIf(accimage is None, 'accimage not available')
    def test_accimage_to_tensor(self):
        trans = transforms.ToTensor()

        expected_output = trans(Image.open(GRACE_HOPPER).convert('RGB'))
        output = trans(accimage.Image(GRACE_HOPPER))

        self.assertEqual(expected_output.size(), output.size())
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
        self.assertTrue(np.allclose(output.numpy(), expected_output.numpy()))

    def test_pil_to_tensor(self):
        test_channels = [1, 3, 4]
        height, width = 4, 4
        trans = transforms.PILToTensor()

        with self.assertRaises(TypeError):
            trans(np.random.rand(1, height, width).tolist())
            trans(np.random.rand(1, height, width))

        for channels in test_channels:
            input_data = torch.ByteTensor(channels, height, width).random_(0, 255)
            img = transforms.ToPILImage()(input_data)
            output = trans(img)
            self.assertTrue(np.allclose(input_data.numpy(), output.numpy()))

            input_data = np.random.randint(low=0, high=255, size=(height, width, channels)).astype(np.uint8)
            img = transforms.ToPILImage()(input_data)
            output = trans(img)
            expected_output = input_data.transpose((2, 0, 1))
            self.assertTrue(np.allclose(output.numpy(), expected_output))

            input_data = torch.as_tensor(np.random.rand(channels, height, width).astype(np.float32))
            img = transforms.ToPILImage()(input_data)  # CHW -> HWC and (* 255).byte()
            output = trans(img)  # HWC -> CHW
            expected_output = (input_data * 255).byte()
            self.assertTrue(np.allclose(output.numpy(), expected_output.numpy()))

        # separate test for mode '1' PIL images
        input_data = torch.ByteTensor(1, height, width).bernoulli_()
        img = transforms.ToPILImage()(input_data.mul(255)).convert('1')
        output = trans(img)
        self.assertTrue(np.allclose(input_data.numpy(), output.numpy()))

    @unittest.skipIf(accimage is None, 'accimage not available')
    def test_accimage_pil_to_tensor(self):
        trans = transforms.PILToTensor()

        expected_output = trans(Image.open(GRACE_HOPPER).convert('RGB'))
        output = trans(accimage.Image(GRACE_HOPPER))

        self.assertEqual(expected_output.size(), output.size())
823
        self.assertTrue(np.allclose(output.numpy(), expected_output.numpy()))
824
825
826
827

    @unittest.skipIf(accimage is None, 'accimage not available')
    def test_accimage_resize(self):
        trans = transforms.Compose([
828
            transforms.Resize(256, interpolation=Image.LINEAR),
829
830
831
            transforms.ToTensor(),
        ])

832
833
834
        # Checking if Compose, Resize and ToTensor can be printed as string
        trans.__repr__()

835
836
837
838
839
840
841
        expected_output = trans(Image.open(GRACE_HOPPER).convert('RGB'))
        output = trans(accimage.Image(GRACE_HOPPER))

        self.assertEqual(expected_output.size(), output.size())
        self.assertLess(np.abs((expected_output - output).mean()), 1e-3)
        self.assertLess((expected_output - output).var(), 1e-5)
        # note the high absolute tolerance
842
        self.assertTrue(np.allclose(output.numpy(), expected_output.numpy(), atol=5e-2))
843
844
845
846
847
848
849
850

    @unittest.skipIf(accimage is None, 'accimage not available')
    def test_accimage_crop(self):
        trans = transforms.Compose([
            transforms.CenterCrop(256),
            transforms.ToTensor(),
        ])

851
852
853
        # Checking if Compose, CenterCrop and ToTensor can be printed as string
        trans.__repr__()

854
855
856
857
        expected_output = trans(Image.open(GRACE_HOPPER).convert('RGB'))
        output = trans(accimage.Image(GRACE_HOPPER))

        self.assertEqual(expected_output.size(), output.size())
858
        self.assertTrue(np.allclose(output.numpy(), expected_output.numpy()))
859

860
    def test_1_channel_tensor_to_pil_image(self):
861
862
        to_tensor = transforms.ToTensor()

863
        img_data_float = torch.Tensor(1, 4, 4).uniform_()
864
865
866
867
        img_data_byte = torch.ByteTensor(1, 4, 4).random_(0, 255)
        img_data_short = torch.ShortTensor(1, 4, 4).random_()
        img_data_int = torch.IntTensor(1, 4, 4).random_()

868
869
870
871
872
873
874
875
876
877
        inputs = [img_data_float, img_data_byte, img_data_short, img_data_int]
        expected_outputs = [img_data_float.mul(255).int().float().div(255).numpy(),
                            img_data_byte.float().div(255.0).numpy(),
                            img_data_short.numpy(),
                            img_data_int.numpy()]
        expected_modes = ['L', 'L', 'I;16', 'I']

        for img_data, expected_output, mode in zip(inputs, expected_outputs, expected_modes):
            for transform in [transforms.ToPILImage(), transforms.ToPILImage(mode=mode)]:
                img = transform(img_data)
878
879
                self.assertEqual(img.mode, mode)
                self.assertTrue(np.allclose(expected_output, to_tensor(img).numpy()))
880
881
        # 'F' mode for torch.FloatTensor
        img_F_mode = transforms.ToPILImage(mode='F')(img_data_float)
882
883
884
        self.assertEqual(img_F_mode.mode, 'F')
        self.assertTrue(np.allclose(np.array(Image.fromarray(img_data_float.squeeze(0).numpy(), mode='F')),
                                    np.array(img_F_mode)))
885
886
887
888
889
890
891
892
893
894
895
896

    def test_1_channel_ndarray_to_pil_image(self):
        img_data_float = torch.Tensor(4, 4, 1).uniform_().numpy()
        img_data_byte = torch.ByteTensor(4, 4, 1).random_(0, 255).numpy()
        img_data_short = torch.ShortTensor(4, 4, 1).random_().numpy()
        img_data_int = torch.IntTensor(4, 4, 1).random_().numpy()

        inputs = [img_data_float, img_data_byte, img_data_short, img_data_int]
        expected_modes = ['F', 'L', 'I;16', 'I']
        for img_data, mode in zip(inputs, expected_modes):
            for transform in [transforms.ToPILImage(), transforms.ToPILImage(mode=mode)]:
                img = transform(img_data)
897
898
                self.assertEqual(img.mode, mode)
                self.assertTrue(np.allclose(img_data[:, :, 0], img))
899

surgan12's avatar
surgan12 committed
900
901
902
903
    def test_2_channel_ndarray_to_pil_image(self):
        def verify_img_data(img_data, mode):
            if mode is None:
                img = transforms.ToPILImage()(img_data)
904
                self.assertEqual(img.mode, 'LA')  # default should assume LA
surgan12's avatar
surgan12 committed
905
906
            else:
                img = transforms.ToPILImage(mode=mode)(img_data)
907
                self.assertEqual(img.mode, mode)
surgan12's avatar
surgan12 committed
908
909
            split = img.split()
            for i in range(2):
910
                self.assertTrue(np.allclose(img_data[:, :, i], split[i]))
surgan12's avatar
surgan12 committed
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927

        img_data = torch.ByteTensor(4, 4, 2).random_(0, 255).numpy()
        for mode in [None, 'LA']:
            verify_img_data(img_data, mode)

        transforms.ToPILImage().__repr__()

        with self.assertRaises(ValueError):
            # should raise if we try a mode for 4 or 1 or 3 channel images
            transforms.ToPILImage(mode='RGBA')(img_data)
            transforms.ToPILImage(mode='P')(img_data)
            transforms.ToPILImage(mode='RGB')(img_data)

    def test_2_channel_tensor_to_pil_image(self):
        def verify_img_data(img_data, expected_output, mode):
            if mode is None:
                img = transforms.ToPILImage()(img_data)
928
                self.assertEqual(img.mode, 'LA')  # default should assume LA
surgan12's avatar
surgan12 committed
929
930
            else:
                img = transforms.ToPILImage(mode=mode)(img_data)
931
                self.assertEqual(img.mode, mode)
surgan12's avatar
surgan12 committed
932
933
            split = img.split()
            for i in range(2):
934
                self.assertTrue(np.allclose(expected_output[i].numpy(), F.to_tensor(split[i]).numpy()))
surgan12's avatar
surgan12 committed
935
936
937
938
939
940
941
942
943
944
945
946

        img_data = torch.Tensor(2, 4, 4).uniform_()
        expected_output = img_data.mul(255).int().float().div(255)
        for mode in [None, 'LA']:
            verify_img_data(img_data, expected_output, mode=mode)

        with self.assertRaises(ValueError):
            # should raise if we try a mode for 4 or 1 or 3 channel images
            transforms.ToPILImage(mode='RGBA')(img_data)
            transforms.ToPILImage(mode='P')(img_data)
            transforms.ToPILImage(mode='RGB')(img_data)

947
948
949
950
    def test_3_channel_tensor_to_pil_image(self):
        def verify_img_data(img_data, expected_output, mode):
            if mode is None:
                img = transforms.ToPILImage()(img_data)
951
                self.assertEqual(img.mode, 'RGB')  # default should assume RGB
952
953
            else:
                img = transforms.ToPILImage(mode=mode)(img_data)
954
                self.assertEqual(img.mode, mode)
955
956
            split = img.split()
            for i in range(3):
957
                self.assertTrue(np.allclose(expected_output[i].numpy(), F.to_tensor(split[i]).numpy()))
958

959
960
961
962
        img_data = torch.Tensor(3, 4, 4).uniform_()
        expected_output = img_data.mul(255).int().float().div(255)
        for mode in [None, 'RGB', 'HSV', 'YCbCr']:
            verify_img_data(img_data, expected_output, mode=mode)
963

964
        with self.assertRaises(ValueError):
surgan12's avatar
surgan12 committed
965
            # should raise if we try a mode for 4 or 1 or 2 channel images
966
967
            transforms.ToPILImage(mode='RGBA')(img_data)
            transforms.ToPILImage(mode='P')(img_data)
surgan12's avatar
surgan12 committed
968
            transforms.ToPILImage(mode='LA')(img_data)
969

Varun Agrawal's avatar
Varun Agrawal committed
970
971
972
        with self.assertRaises(ValueError):
            transforms.ToPILImage()(torch.Tensor(1, 3, 4, 4).uniform_())

973
974
975
976
    def test_3_channel_ndarray_to_pil_image(self):
        def verify_img_data(img_data, mode):
            if mode is None:
                img = transforms.ToPILImage()(img_data)
977
                self.assertEqual(img.mode, 'RGB')  # default should assume RGB
978
979
            else:
                img = transforms.ToPILImage(mode=mode)(img_data)
980
                self.assertEqual(img.mode, mode)
981
982
            split = img.split()
            for i in range(3):
983
                self.assertTrue(np.allclose(img_data[:, :, i], split[i]))
984

985
986
987
988
        img_data = torch.ByteTensor(4, 4, 3).random_(0, 255).numpy()
        for mode in [None, 'RGB', 'HSV', 'YCbCr']:
            verify_img_data(img_data, mode)

989
990
991
        # Checking if ToPILImage can be printed as string
        transforms.ToPILImage().__repr__()

992
        with self.assertRaises(ValueError):
surgan12's avatar
surgan12 committed
993
            # should raise if we try a mode for 4 or 1 or 2 channel images
994
995
            transforms.ToPILImage(mode='RGBA')(img_data)
            transforms.ToPILImage(mode='P')(img_data)
surgan12's avatar
surgan12 committed
996
            transforms.ToPILImage(mode='LA')(img_data)
997
998
999
1000
1001

    def test_4_channel_tensor_to_pil_image(self):
        def verify_img_data(img_data, expected_output, mode):
            if mode is None:
                img = transforms.ToPILImage()(img_data)
1002
                self.assertEqual(img.mode, 'RGBA')  # default should assume RGBA
1003
1004
            else:
                img = transforms.ToPILImage(mode=mode)(img_data)
1005
                self.assertEqual(img.mode, mode)
1006
1007
1008

            split = img.split()
            for i in range(4):
1009
                self.assertTrue(np.allclose(expected_output[i].numpy(), F.to_tensor(split[i]).numpy()))
1010

1011
        img_data = torch.Tensor(4, 4, 4).uniform_()
1012
        expected_output = img_data.mul(255).int().float().div(255)
surgan12's avatar
surgan12 committed
1013
        for mode in [None, 'RGBA', 'CMYK', 'RGBX']:
1014
            verify_img_data(img_data, expected_output, mode)
1015

1016
        with self.assertRaises(ValueError):
surgan12's avatar
surgan12 committed
1017
            # should raise if we try a mode for 3 or 1 or 2 channel images
1018
1019
            transforms.ToPILImage(mode='RGB')(img_data)
            transforms.ToPILImage(mode='P')(img_data)
surgan12's avatar
surgan12 committed
1020
            transforms.ToPILImage(mode='LA')(img_data)
1021
1022
1023
1024
1025

    def test_4_channel_ndarray_to_pil_image(self):
        def verify_img_data(img_data, mode):
            if mode is None:
                img = transforms.ToPILImage()(img_data)
1026
                self.assertEqual(img.mode, 'RGBA')  # default should assume RGBA
1027
1028
            else:
                img = transforms.ToPILImage(mode=mode)(img_data)
1029
                self.assertEqual(img.mode, mode)
1030
1031
            split = img.split()
            for i in range(4):
1032
                self.assertTrue(np.allclose(img_data[:, :, i], split[i]))
1033

1034
        img_data = torch.ByteTensor(4, 4, 4).random_(0, 255).numpy()
surgan12's avatar
surgan12 committed
1035
        for mode in [None, 'RGBA', 'CMYK', 'RGBX']:
1036
            verify_img_data(img_data, mode)
1037

1038
        with self.assertRaises(ValueError):
surgan12's avatar
surgan12 committed
1039
            # should raise if we try a mode for 3 or 1 or 2 channel images
1040
1041
            transforms.ToPILImage(mode='RGB')(img_data)
            transforms.ToPILImage(mode='P')(img_data)
surgan12's avatar
surgan12 committed
1042
            transforms.ToPILImage(mode='LA')(img_data)
1043

Varun Agrawal's avatar
Varun Agrawal committed
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
    def test_2d_tensor_to_pil_image(self):
        to_tensor = transforms.ToTensor()

        img_data_float = torch.Tensor(4, 4).uniform_()
        img_data_byte = torch.ByteTensor(4, 4).random_(0, 255)
        img_data_short = torch.ShortTensor(4, 4).random_()
        img_data_int = torch.IntTensor(4, 4).random_()

        inputs = [img_data_float, img_data_byte, img_data_short, img_data_int]
        expected_outputs = [img_data_float.mul(255).int().float().div(255).numpy(),
                            img_data_byte.float().div(255.0).numpy(),
                            img_data_short.numpy(),
                            img_data_int.numpy()]
        expected_modes = ['L', 'L', 'I;16', 'I']

        for img_data, expected_output, mode in zip(inputs, expected_outputs, expected_modes):
            for transform in [transforms.ToPILImage(), transforms.ToPILImage(mode=mode)]:
                img = transform(img_data)
1062
1063
                self.assertEqual(img.mode, mode)
                self.assertTrue(np.allclose(expected_output, to_tensor(img).numpy()))
Varun Agrawal's avatar
Varun Agrawal committed
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075

    def test_2d_ndarray_to_pil_image(self):
        img_data_float = torch.Tensor(4, 4).uniform_().numpy()
        img_data_byte = torch.ByteTensor(4, 4).random_(0, 255).numpy()
        img_data_short = torch.ShortTensor(4, 4).random_().numpy()
        img_data_int = torch.IntTensor(4, 4).random_().numpy()

        inputs = [img_data_float, img_data_byte, img_data_short, img_data_int]
        expected_modes = ['F', 'L', 'I;16', 'I']
        for img_data, mode in zip(inputs, expected_modes):
            for transform in [transforms.ToPILImage(), transforms.ToPILImage(mode=mode)]:
                img = transform(img_data)
1076
1077
                self.assertEqual(img.mode, mode)
                self.assertTrue(np.allclose(img_data, img))
Varun Agrawal's avatar
Varun Agrawal committed
1078
1079

    def test_tensor_bad_types_to_pil_image(self):
1080
        with self.assertRaisesRegex(ValueError, r'pic should be 2/3 dimensional. Got \d+ dimensions.'):
Varun Agrawal's avatar
Varun Agrawal committed
1081
            transforms.ToPILImage()(torch.ones(1, 3, 4, 4))
1082
1083
        with self.assertRaisesRegex(ValueError, r'pic should not have > 4 channels. Got \d+ channels.'):
            transforms.ToPILImage()(torch.ones(6, 4, 4))
Varun Agrawal's avatar
Varun Agrawal committed
1084

1085
    def test_ndarray_bad_types_to_pil_image(self):
1086
        trans = transforms.ToPILImage()
1087
1088
        reg_msg = r'Input type \w+ is not supported'
        with self.assertRaisesRegex(TypeError, reg_msg):
1089
            trans(np.ones([4, 4, 1], np.int64))
1090
        with self.assertRaisesRegex(TypeError, reg_msg):
1091
            trans(np.ones([4, 4, 1], np.uint16))
1092
        with self.assertRaisesRegex(TypeError, reg_msg):
1093
            trans(np.ones([4, 4, 1], np.uint32))
1094
        with self.assertRaisesRegex(TypeError, reg_msg):
1095
1096
            trans(np.ones([4, 4, 1], np.float64))

1097
        with self.assertRaisesRegex(ValueError, r'pic should be 2/3 dimensional. Got \d+ dimensions.'):
Varun Agrawal's avatar
Varun Agrawal committed
1098
            transforms.ToPILImage()(np.ones([1, 4, 4, 3]))
1099
1100
        with self.assertRaisesRegex(ValueError, r'pic should not have > 4 channels. Got \d+ channels.'):
            transforms.ToPILImage()(np.ones([4, 4, 6]))
Varun Agrawal's avatar
Varun Agrawal committed
1101

1102
1103
    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_vertical_flip(self):
1104
1105
        random_state = random.getstate()
        random.seed(42)
1106
1107
1108
        img = transforms.ToPILImage()(torch.rand(3, 10, 10))
        vimg = img.transpose(Image.FLIP_TOP_BOTTOM)

1109
        num_samples = 250
1110
        num_vertical = 0
1111
        for _ in range(num_samples):
1112
1113
1114
1115
            out = transforms.RandomVerticalFlip()(img)
            if out == vimg:
                num_vertical += 1

1116
1117
        p_value = stats.binom_test(num_vertical, num_samples, p=0.5)
        random.setstate(random_state)
1118
        self.assertGreater(p_value, 0.0001)
1119

1120
1121
1122
1123
1124
1125
1126
1127
1128
        num_samples = 250
        num_vertical = 0
        for _ in range(num_samples):
            out = transforms.RandomVerticalFlip(p=0.7)(img)
            if out == vimg:
                num_vertical += 1

        p_value = stats.binom_test(num_vertical, num_samples, p=0.7)
        random.setstate(random_state)
1129
        self.assertGreater(p_value, 0.0001)
1130

1131
1132
1133
        # Checking if RandomVerticalFlip can be printed as string
        transforms.RandomVerticalFlip().__repr__()

1134
1135
    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_horizontal_flip(self):
1136
1137
        random_state = random.getstate()
        random.seed(42)
1138
1139
1140
        img = transforms.ToPILImage()(torch.rand(3, 10, 10))
        himg = img.transpose(Image.FLIP_LEFT_RIGHT)

1141
        num_samples = 250
1142
        num_horizontal = 0
1143
        for _ in range(num_samples):
1144
1145
1146
1147
            out = transforms.RandomHorizontalFlip()(img)
            if out == himg:
                num_horizontal += 1

1148
1149
        p_value = stats.binom_test(num_horizontal, num_samples, p=0.5)
        random.setstate(random_state)
1150
        self.assertGreater(p_value, 0.0001)
1151

1152
1153
1154
1155
1156
1157
1158
1159
1160
        num_samples = 250
        num_horizontal = 0
        for _ in range(num_samples):
            out = transforms.RandomHorizontalFlip(p=0.7)(img)
            if out == himg:
                num_horizontal += 1

        p_value = stats.binom_test(num_horizontal, num_samples, p=0.7)
        random.setstate(random_state)
1161
        self.assertGreater(p_value, 0.0001)
1162

1163
1164
1165
        # Checking if RandomHorizontalFlip can be printed as string
        transforms.RandomHorizontalFlip().__repr__()

1166
    @unittest.skipIf(stats is None, 'scipy.stats is not available')
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
    def test_normalize(self):
        def samples_from_standard_normal(tensor):
            p_value = stats.kstest(list(tensor.view(-1)), 'norm', args=(0, 1)).pvalue
            return p_value > 0.0001

        random_state = random.getstate()
        random.seed(42)
        for channels in [1, 3]:
            img = torch.rand(channels, 10, 10)
            mean = [img[c].mean() for c in range(channels)]
            std = [img[c].std() for c in range(channels)]
            normalized = transforms.Normalize(mean, std)(img)
1179
            self.assertTrue(samples_from_standard_normal(normalized))
1180
1181
        random.setstate(random_state)

1182
1183
1184
        # Checking if Normalize can be printed as string
        transforms.Normalize(mean, std).__repr__()

1185
1186
1187
        # Checking the optional in-place behaviour
        tensor = torch.rand((1, 16, 16))
        tensor_inplace = transforms.Normalize((0.5,), (0.5,), inplace=True)(tensor)
1188
        self.assertTrue(torch.equal(tensor, tensor_inplace))
1189

1190
1191
1192
1193
1194
1195
1196
1197
1198
    def test_normalize_different_dtype(self):
        for dtype1 in [torch.float32, torch.float64]:
            img = torch.rand(3, 10, 10, dtype=dtype1)
            for dtype2 in [torch.int64, torch.float32, torch.float64]:
                mean = torch.tensor([1, 2, 3], dtype=dtype2)
                std = torch.tensor([1, 2, 1], dtype=dtype2)
                # checks that it doesn't crash
                transforms.functional.normalize(img, mean, std)

1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
    def test_normalize_3d_tensor(self):
        torch.manual_seed(28)
        n_channels = 3
        img_size = 10
        mean = torch.rand(n_channels)
        std = torch.rand(n_channels)
        img = torch.rand(n_channels, img_size, img_size)
        target = F.normalize(img, mean, std).numpy()

        mean_unsqueezed = mean.view(-1, 1, 1)
        std_unsqueezed = std.view(-1, 1, 1)
        result1 = F.normalize(img, mean_unsqueezed, std_unsqueezed)
        result2 = F.normalize(img,
                              mean_unsqueezed.repeat(1, img_size, img_size),
                              std_unsqueezed.repeat(1, img_size, img_size))
        assert_array_almost_equal(target, result1.numpy())
        assert_array_almost_equal(target, result2.numpy())

1217
1218
1219
1220
1221
1222
1223
    def test_adjust_brightness(self):
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')

        # test 0
1224
        y_pil = F.adjust_brightness(x_pil, 1)
1225
        y_np = np.array(y_pil)
1226
        self.assertTrue(np.allclose(y_np, x_np))
1227
1228

        # test 1
1229
        y_pil = F.adjust_brightness(x_pil, 0.5)
1230
1231
1232
        y_np = np.array(y_pil)
        y_ans = [0, 2, 6, 27, 67, 113, 18, 4, 117, 45, 127, 0]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1233
        self.assertTrue(np.allclose(y_np, y_ans))
1234
1235

        # test 2
1236
        y_pil = F.adjust_brightness(x_pil, 2)
1237
1238
1239
        y_np = np.array(y_pil)
        y_ans = [0, 10, 26, 108, 255, 255, 74, 16, 255, 180, 255, 2]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1240
        self.assertTrue(np.allclose(y_np, y_ans))
1241
1242
1243
1244
1245
1246
1247
1248

    def test_adjust_contrast(self):
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')

        # test 0
1249
        y_pil = F.adjust_contrast(x_pil, 1)
1250
        y_np = np.array(y_pil)
1251
        self.assertTrue(np.allclose(y_np, x_np))
1252
1253

        # test 1
1254
        y_pil = F.adjust_contrast(x_pil, 0.5)
1255
1256
1257
        y_np = np.array(y_pil)
        y_ans = [43, 45, 49, 70, 110, 156, 61, 47, 160, 88, 170, 43]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1258
        self.assertTrue(np.allclose(y_np, y_ans))
1259
1260

        # test 2
1261
        y_pil = F.adjust_contrast(x_pil, 2)
1262
1263
1264
        y_np = np.array(y_pil)
        y_ans = [0, 0, 0, 22, 184, 255, 0, 0, 255, 94, 255, 0]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1265
        self.assertTrue(np.allclose(y_np, y_ans))
1266

Francisco Massa's avatar
Francisco Massa committed
1267
    @unittest.skipIf(Image.__version__ >= '7', "Temporarily disabled")
1268
1269
1270
1271
1272
1273
1274
    def test_adjust_saturation(self):
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')

        # test 0
1275
        y_pil = F.adjust_saturation(x_pil, 1)
1276
        y_np = np.array(y_pil)
1277
        self.assertTrue(np.allclose(y_np, x_np))
1278
1279

        # test 1
1280
        y_pil = F.adjust_saturation(x_pil, 0.5)
1281
1282
1283
        y_np = np.array(y_pil)
        y_ans = [2, 4, 8, 87, 128, 173, 39, 25, 138, 133, 215, 88]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1284
        self.assertTrue(np.allclose(y_np, y_ans))
1285
1286

        # test 2
1287
        y_pil = F.adjust_saturation(x_pil, 2)
1288
1289
1290
        y_np = np.array(y_pil)
        y_ans = [0, 6, 22, 0, 149, 255, 32, 0, 255, 4, 255, 0]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1291
        self.assertTrue(np.allclose(y_np, y_ans))
1292
1293
1294
1295
1296
1297
1298
1299

    def test_adjust_hue(self):
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')

        with self.assertRaises(ValueError):
1300
1301
            F.adjust_hue(x_pil, -0.7)
            F.adjust_hue(x_pil, 1)
1302
1303
1304

        # test 0: almost same as x_data but not exact.
        # probably because hsv <-> rgb floating point ops
1305
        y_pil = F.adjust_hue(x_pil, 0)
1306
1307
1308
        y_np = np.array(y_pil)
        y_ans = [0, 5, 13, 54, 139, 226, 35, 8, 234, 91, 255, 1]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1309
        self.assertTrue(np.allclose(y_np, y_ans))
1310
1311

        # test 1
1312
        y_pil = F.adjust_hue(x_pil, 0.25)
1313
1314
1315
        y_np = np.array(y_pil)
        y_ans = [13, 0, 12, 224, 54, 226, 234, 8, 99, 1, 222, 255]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1316
        self.assertTrue(np.allclose(y_np, y_ans))
1317
1318

        # test 2
1319
        y_pil = F.adjust_hue(x_pil, -0.25)
1320
1321
1322
        y_np = np.array(y_pil)
        y_ans = [0, 13, 2, 54, 226, 58, 8, 234, 152, 255, 43, 1]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1323
        self.assertTrue(np.allclose(y_np, y_ans))
1324

1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
    def test_adjust_sharpness(self):
        x_shape = [4, 4, 3]
        x_data = [75, 121, 114, 105, 97, 107, 105, 32, 66, 111, 117, 114, 99, 104, 97, 0,
                  0, 65, 108, 101, 120, 97, 110, 100, 101, 114, 32, 86, 114, 121, 110, 105,
                  111, 116, 105, 115, 0, 0, 73, 32, 108, 111, 118, 101, 32, 121, 111, 117]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')

        # test 0
        y_pil = F.adjust_sharpness(x_pil, 1)
        y_np = np.array(y_pil)
        self.assertTrue(np.allclose(y_np, x_np))

        # test 1
        y_pil = F.adjust_sharpness(x_pil, 0.5)
        y_np = np.array(y_pil)
        y_ans = [75, 121, 114, 105, 97, 107, 105, 32, 66, 111, 117, 114, 99, 104, 97, 30,
                 30, 74, 103, 96, 114, 97, 110, 100, 101, 114, 32, 81, 103, 108, 102, 101,
                 107, 116, 105, 115, 0, 0, 73, 32, 108, 111, 118, 101, 32, 121, 111, 117]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
        self.assertTrue(np.allclose(y_np, y_ans))

        # test 2
        y_pil = F.adjust_sharpness(x_pil, 2)
        y_np = np.array(y_pil)
        y_ans = [75, 121, 114, 105, 97, 107, 105, 32, 66, 111, 117, 114, 99, 104, 97, 0,
                 0, 46, 118, 111, 132, 97, 110, 100, 101, 114, 32, 95, 135, 146, 126, 112,
                 119, 116, 105, 115, 0, 0, 73, 32, 108, 111, 118, 101, 32, 121, 111, 117]
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
        self.assertTrue(np.allclose(y_np, y_ans))

        # test 3
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')
        x_th = torch.tensor(x_np.transpose(2, 0, 1))
        y_pil = F.adjust_sharpness(x_pil, 2)
        y_np = np.array(y_pil).transpose(2, 0, 1)
        y_th = F.adjust_sharpness(x_th, 2)
        self.assertTrue(np.allclose(y_np, y_th.numpy()))

1367
1368
1369
1370
1371
1372
1373
    def test_adjust_gamma(self):
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')

        # test 0
1374
        y_pil = F.adjust_gamma(x_pil, 1)
1375
        y_np = np.array(y_pil)
1376
        self.assertTrue(np.allclose(y_np, x_np))
1377
1378

        # test 1
1379
        y_pil = F.adjust_gamma(x_pil, 0.5)
1380
        y_np = np.array(y_pil)
1381
        y_ans = [0, 35, 57, 117, 186, 241, 97, 45, 245, 152, 255, 16]
1382
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1383
        self.assertTrue(np.allclose(y_np, y_ans))
1384
1385

        # test 2
1386
        y_pil = F.adjust_gamma(x_pil, 2)
1387
        y_np = np.array(y_pil)
1388
        y_ans = [0, 0, 0, 11, 71, 201, 5, 0, 215, 31, 255, 0]
1389
        y_ans = np.array(y_ans, dtype=np.uint8).reshape(x_shape)
1390
        self.assertTrue(np.allclose(y_np, y_ans))
1391
1392
1393
1394
1395
1396
1397
1398

    def test_adjusts_L_mode(self):
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_rgb = Image.fromarray(x_np, mode='RGB')

        x_l = x_rgb.convert('L')
1399
1400
1401
1402
        self.assertEqual(F.adjust_brightness(x_l, 2).mode, 'L')
        self.assertEqual(F.adjust_saturation(x_l, 2).mode, 'L')
        self.assertEqual(F.adjust_contrast(x_l, 2).mode, 'L')
        self.assertEqual(F.adjust_hue(x_l, 0.4).mode, 'L')
1403
        self.assertEqual(F.adjust_sharpness(x_l, 2).mode, 'L')
1404
        self.assertEqual(F.adjust_gamma(x_l, 0.5).mode, 'L')
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416

    def test_color_jitter(self):
        color_jitter = transforms.ColorJitter(2, 2, 2, 0.1)

        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')
        x_pil_2 = x_pil.convert('L')

        for i in range(10):
            y_pil = color_jitter(x_pil)
1417
            self.assertEqual(y_pil.mode, x_pil.mode)
1418
1419

            y_pil_2 = color_jitter(x_pil_2)
1420
            self.assertEqual(y_pil_2.mode, x_pil_2.mode)
1421

1422
1423
1424
        # Checking if ColorJitter can be printed as string
        color_jitter.__repr__()

1425
    def test_linear_transformation(self):
ekka's avatar
ekka committed
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
        num_samples = 1000
        x = torch.randn(num_samples, 3, 10, 10)
        flat_x = x.view(x.size(0), x.size(1) * x.size(2) * x.size(3))
        # compute principal components
        sigma = torch.mm(flat_x.t(), flat_x) / flat_x.size(0)
        u, s, _ = np.linalg.svd(sigma.numpy())
        zca_epsilon = 1e-10  # avoid division by 0
        d = torch.Tensor(np.diag(1. / np.sqrt(s + zca_epsilon)))
        u = torch.Tensor(u)
        principal_components = torch.mm(torch.mm(u, d), u.t())
        mean_vector = (torch.sum(flat_x, dim=0) / flat_x.size(0))
        # initialize whitening matrix
1438
        whitening = transforms.LinearTransformation(principal_components, mean_vector)
ekka's avatar
ekka committed
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
        # estimate covariance and mean using weak law of large number
        num_features = flat_x.size(1)
        cov = 0.0
        mean = 0.0
        for i in x:
            xwhite = whitening(i)
            xwhite = xwhite.view(1, -1).numpy()
            cov += np.dot(xwhite, xwhite.T) / num_features
            mean += np.sum(xwhite) / num_features
        # if rtol for std = 1e-3 then rtol for cov = 2e-3 as std**2 = cov
1449
1450
1451
1452
        self.assertTrue(np.allclose(cov / num_samples, np.identity(1), rtol=2e-3),
                        "cov not close to 1")
        self.assertTrue(np.allclose(mean / num_samples, 0, rtol=1e-3),
                        "mean not close to 0")
ekka's avatar
ekka committed
1453

1454
        # Checking if LinearTransformation can be printed as string
ekka's avatar
ekka committed
1455
1456
        whitening.__repr__()

1457
1458
1459
1460
    def test_rotate(self):
        x = np.zeros((100, 100, 3), dtype=np.uint8)
        x[40, 40] = [255, 255, 255]

vfdev's avatar
vfdev committed
1461
        with self.assertRaisesRegex(TypeError, r"img should be PIL Image"):
1462
1463
1464
1465
1466
            F.rotate(x, 10)

        img = F.to_pil_image(x)

        result = F.rotate(img, 45)
1467
        self.assertEqual(result.size, (100, 100))
1468
        r, c, ch = np.where(result)
1469
1470
1471
        self.assertTrue(all(x in r for x in [49, 50]))
        self.assertTrue(all(x in c for x in [36]))
        self.assertTrue(all(x in ch for x in [0, 1, 2]))
1472
1473

        result = F.rotate(img, 45, expand=True)
1474
        self.assertEqual(result.size, (142, 142))
1475
        r, c, ch = np.where(result)
1476
1477
1478
        self.assertTrue(all(x in r for x in [70, 71]))
        self.assertTrue(all(x in c for x in [57]))
        self.assertTrue(all(x in ch for x in [0, 1, 2]))
1479
1480

        result = F.rotate(img, 45, center=(40, 40))
1481
        self.assertEqual(result.size, (100, 100))
1482
        r, c, ch = np.where(result)
1483
1484
1485
        self.assertTrue(all(x in r for x in [40]))
        self.assertTrue(all(x in c for x in [40]))
        self.assertTrue(all(x in ch for x in [0, 1, 2]))
1486
1487
1488
1489

        result_a = F.rotate(img, 90)
        result_b = F.rotate(img, -270)

1490
        self.assertTrue(np.all(np.array(result_a) == np.array(result_b)))
1491

Philip Meier's avatar
Philip Meier committed
1492
1493
1494
    def test_rotate_fill(self):
        img = F.to_pil_image(np.ones((100, 100, 3), dtype=np.uint8) * 255, "RGB")

1495
        modes = ("L", "RGB", "F")
Philip Meier's avatar
Philip Meier committed
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
        nums_bands = [len(mode) for mode in modes]
        fill = 127

        for mode, num_bands in zip(modes, nums_bands):
            img_conv = img.convert(mode)
            img_rot = F.rotate(img_conv, 45.0, fill=fill)
            pixel = img_rot.getpixel((0, 0))

            if not isinstance(pixel, tuple):
                pixel = (pixel,)
            self.assertTupleEqual(pixel, tuple([fill] * num_bands))

            for wrong_num_bands in set(nums_bands) - {num_bands}:
                with self.assertRaises(ValueError):
                    F.rotate(img_conv, 45.0, fill=tuple([fill] * wrong_num_bands))

1512
    def test_affine(self):
Francisco Massa's avatar
Francisco Massa committed
1513
1514
1515
        input_img = np.zeros((40, 40, 3), dtype=np.uint8)
        cnt = [20, 20]
        for pt in [(16, 16), (20, 16), (20, 20)]:
1516
1517
1518
1519
            for i in range(-5, 5):
                for j in range(-5, 5):
                    input_img[pt[0] + i, pt[1] + j, :] = [255, 155, 55]

vfdev's avatar
vfdev committed
1520
1521
        with self.assertRaises(TypeError, msg="Argument translate should be a sequence"):
            F.affine(input_img, 10, translate=0, scale=1, shear=1)
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532

        pil_img = F.to_pil_image(input_img)

        def _to_3x3_inv(inv_result_matrix):
            result_matrix = np.zeros((3, 3))
            result_matrix[:2, :] = np.array(inv_result_matrix).reshape((2, 3))
            result_matrix[2, 2] = 1
            return np.linalg.inv(result_matrix)

        def _test_transformation(a, t, s, sh):
            a_rad = math.radians(a)
ptrblck's avatar
ptrblck committed
1533
            s_rad = [math.radians(sh_) for sh_ in sh]
1534
1535
1536
1537
1538
            cx, cy = cnt
            tx, ty = t
            sx, sy = s_rad
            rot = a_rad

1539
            # 1) Check transformation matrix:
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
            C = np.array([[1, 0, cx],
                          [0, 1, cy],
                          [0, 0, 1]])
            T = np.array([[1, 0, tx],
                          [0, 1, ty],
                          [0, 0, 1]])
            Cinv = np.linalg.inv(C)

            RS = np.array(
                [[s * math.cos(rot), -s * math.sin(rot), 0],
                 [s * math.sin(rot), s * math.cos(rot), 0],
                 [0, 0, 1]])

            SHx = np.array([[1, -math.tan(sx), 0],
                            [0, 1, 0],
                            [0, 0, 1]])

            SHy = np.array([[1, 0, 0],
                            [-math.tan(sy), 1, 0],
                            [0, 0, 1]])

            RSS = np.matmul(RS, np.matmul(SHy, SHx))

            true_matrix = np.matmul(T, np.matmul(C, np.matmul(RSS, Cinv)))

1565
1566
            result_matrix = _to_3x3_inv(F._get_inverse_affine_matrix(center=cnt, angle=a,
                                                                     translate=t, scale=s, shear=sh))
1567
            self.assertLess(np.sum(np.abs(true_matrix - result_matrix)), 1e-10)
1568
            # 2) Perform inverse mapping:
Francisco Massa's avatar
Francisco Massa committed
1569
            true_result = np.zeros((40, 40, 3), dtype=np.uint8)
1570
1571
1572
            inv_true_matrix = np.linalg.inv(true_matrix)
            for y in range(true_result.shape[0]):
                for x in range(true_result.shape[1]):
1573
1574
1575
1576
1577
1578
                    # Same as for PIL:
                    # https://github.com/python-pillow/Pillow/blob/71f8ec6a0cfc1008076a023c0756542539d057ab/
                    # src/libImaging/Geometry.c#L1060
                    input_pt = np.array([x + 0.5, y + 0.5, 1.0])
                    res = np.floor(np.dot(inv_true_matrix, input_pt)).astype(np.int)
                    _x, _y = res[:2]
1579
1580
1581
1582
                    if 0 <= _x < input_img.shape[1] and 0 <= _y < input_img.shape[0]:
                        true_result[y, x, :] = input_img[_y, _x, :]

            result = F.affine(pil_img, angle=a, translate=t, scale=s, shear=sh)
1583
            self.assertEqual(result.size, pil_img.size)
1584
1585
1586
1587
            # Compute number of different pixels:
            np_result = np.array(result)
            n_diff_pixels = np.sum(np_result != true_result) / 3
            # Accept 3 wrong pixels
1588
1589
1590
            self.assertLess(n_diff_pixels, 3,
                            "a={}, t={}, s={}, sh={}\n".format(a, t, s, sh) +
                            "n diff pixels={}\n".format(np.sum(np.array(result)[:, :, 0] != true_result[:, :, 0])))
1591
1592
1593

        # Test rotation
        a = 45
ptrblck's avatar
ptrblck committed
1594
        _test_transformation(a=a, t=(0, 0), s=1.0, sh=(0.0, 0.0))
1595
1596
1597

        # Test translation
        t = [10, 15]
ptrblck's avatar
ptrblck committed
1598
        _test_transformation(a=0.0, t=t, s=1.0, sh=(0.0, 0.0))
1599
1600
1601

        # Test scale
        s = 1.2
ptrblck's avatar
ptrblck committed
1602
        _test_transformation(a=0.0, t=(0.0, 0.0), s=s, sh=(0.0, 0.0))
1603
1604

        # Test shear
ptrblck's avatar
ptrblck committed
1605
        sh = [45.0, 25.0]
1606
1607
1608
1609
1610
        _test_transformation(a=0.0, t=(0.0, 0.0), s=1.0, sh=sh)

        # Test rotation, scale, translation, shear
        for a in range(-90, 90, 25):
            for t1 in range(-10, 10, 5):
1611
                for s in [0.75, 0.98, 1.0, 1.2, 1.4]:
1612
                    for sh in range(-15, 15, 5):
ptrblck's avatar
ptrblck committed
1613
                        _test_transformation(a=a, t=(t1, t1), s=s, sh=(sh, sh))
1614

1615
1616
1617
1618
1619
1620
1621
    def test_random_rotation(self):

        with self.assertRaises(ValueError):
            transforms.RandomRotation(-0.7)
            transforms.RandomRotation([-0.7])
            transforms.RandomRotation([-0.7, 0, 0.7])

1622
1623
1624
1625
1626
1627
1628
        # assert fill being either a Sequence or a Number
        with self.assertRaises(TypeError):
            transforms.RandomRotation(0, fill={})

        t = transforms.RandomRotation(0, fill=None)
        self.assertTrue(t.fill == 0)

1629
1630
        t = transforms.RandomRotation(10)
        angle = t.get_params(t.degrees)
1631
        self.assertTrue(angle > -10 and angle < 10)
1632
1633
1634

        t = transforms.RandomRotation((-10, 10))
        angle = t.get_params(t.degrees)
1635
        self.assertTrue(-10 < angle < 10)
1636

1637
1638
1639
        # Checking if RandomRotation can be printed as string
        t.__repr__()

1640
1641
1642
        # assert deprecation warning and non-BC
        with self.assertWarnsRegex(UserWarning, r"Argument resample is deprecated and will be removed"):
            t = transforms.RandomRotation((-10, 10), resample=2)
1643
            self.assertEqual(t.interpolation, transforms.InterpolationMode.BILINEAR)
1644
1645

        # assert changed type warning
1646
        with self.assertWarnsRegex(UserWarning, r"Argument interpolation should be of type InterpolationMode"):
1647
            t = transforms.RandomRotation((-10, 10), interpolation=2)
1648
            self.assertEqual(t.interpolation, transforms.InterpolationMode.BILINEAR)
1649

1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
    def test_random_affine(self):

        with self.assertRaises(ValueError):
            transforms.RandomAffine(-0.7)
            transforms.RandomAffine([-0.7])
            transforms.RandomAffine([-0.7, 0, 0.7])

            transforms.RandomAffine([-90, 90], translate=2.0)
            transforms.RandomAffine([-90, 90], translate=[-1.0, 1.0])
            transforms.RandomAffine([-90, 90], translate=[-1.0, 0.0, 1.0])

            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[0.0])
            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[-1.0, 1.0])
            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[0.5, -0.5])
            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[0.5, 3.0, -0.5])

            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[0.5, 0.5], shear=-7)
            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[0.5, 0.5], shear=[-10])
            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[0.5, 0.5], shear=[-10, 0, 10])
ptrblck's avatar
ptrblck committed
1669
            transforms.RandomAffine([-90, 90], translate=[0.2, 0.2], scale=[0.5, 0.5], shear=[-10, 0, 10, 0, 10])
1670

1671
1672
1673
1674
1675
1676
1677
        # assert fill being either a Sequence or a Number
        with self.assertRaises(TypeError):
            transforms.RandomAffine(0, fill={})

        t = transforms.RandomAffine(0, fill=None)
        self.assertTrue(t.fill == 0)

1678
1679
1680
        x = np.zeros((100, 100, 3), dtype=np.uint8)
        img = F.to_pil_image(x)

ptrblck's avatar
ptrblck committed
1681
        t = transforms.RandomAffine(10, translate=[0.5, 0.3], scale=[0.7, 1.3], shear=[-10, 10, 20, 40])
1682
1683
1684
        for _ in range(100):
            angle, translations, scale, shear = t.get_params(t.degrees, t.translate, t.scale, t.shear,
                                                             img_size=img.size)
1685
1686
1687
1688
1689
1690
1691
1692
            self.assertTrue(-10 < angle < 10)
            self.assertTrue(-img.size[0] * 0.5 <= translations[0] <= img.size[0] * 0.5,
                            "{} vs {}".format(translations[0], img.size[0] * 0.5))
            self.assertTrue(-img.size[1] * 0.5 <= translations[1] <= img.size[1] * 0.5,
                            "{} vs {}".format(translations[1], img.size[1] * 0.5))
            self.assertTrue(0.7 < scale < 1.3)
            self.assertTrue(-10 < shear[0] < 10)
            self.assertTrue(-20 < shear[1] < 40)
1693
1694
1695
1696

        # Checking if RandomAffine can be printed as string
        t.__repr__()

1697
        t = transforms.RandomAffine(10, interpolation=transforms.InterpolationMode.BILINEAR)
1698
1699
1700
1701
1702
        self.assertIn("bilinear", t.__repr__())

        # assert deprecation warning and non-BC
        with self.assertWarnsRegex(UserWarning, r"Argument resample is deprecated and will be removed"):
            t = transforms.RandomAffine(10, resample=2)
1703
            self.assertEqual(t.interpolation, transforms.InterpolationMode.BILINEAR)
1704
1705
1706
1707
1708
1709

        with self.assertWarnsRegex(UserWarning, r"Argument fillcolor is deprecated and will be removed"):
            t = transforms.RandomAffine(10, fillcolor=10)
            self.assertEqual(t.fill, 10)

        # assert changed type warning
1710
        with self.assertWarnsRegex(UserWarning, r"Argument interpolation should be of type InterpolationMode"):
1711
            t = transforms.RandomAffine(10, interpolation=2)
1712
            self.assertEqual(t.interpolation, transforms.InterpolationMode.BILINEAR)
1713

1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
    def test_to_grayscale(self):
        """Unit tests for grayscale transform"""

        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')
        x_pil_2 = x_pil.convert('L')
        gray_np = np.array(x_pil_2)

        # Test Set: Grayscale an image with desired number of output channels
        # Case 1: RGB -> 1 channel grayscale
        trans1 = transforms.Grayscale(num_output_channels=1)
        gray_pil_1 = trans1(x_pil)
        gray_np_1 = np.array(gray_pil_1)
1729
1730
        self.assertEqual(gray_pil_1.mode, 'L', 'mode should be L')
        self.assertEqual(gray_np_1.shape, tuple(x_shape[0:2]), 'should be 1 channel')
1731
1732
1733
1734
1735
1736
        np.testing.assert_equal(gray_np, gray_np_1)

        # Case 2: RGB -> 3 channel grayscale
        trans2 = transforms.Grayscale(num_output_channels=3)
        gray_pil_2 = trans2(x_pil)
        gray_np_2 = np.array(gray_pil_2)
1737
1738
        self.assertEqual(gray_pil_2.mode, 'RGB', 'mode should be RGB')
        self.assertEqual(gray_np_2.shape, tuple(x_shape), 'should be 3 channel')
1739
1740
1741
1742
1743
1744
1745
1746
        np.testing.assert_equal(gray_np_2[:, :, 0], gray_np_2[:, :, 1])
        np.testing.assert_equal(gray_np_2[:, :, 1], gray_np_2[:, :, 2])
        np.testing.assert_equal(gray_np, gray_np_2[:, :, 0])

        # Case 3: 1 channel grayscale -> 1 channel grayscale
        trans3 = transforms.Grayscale(num_output_channels=1)
        gray_pil_3 = trans3(x_pil_2)
        gray_np_3 = np.array(gray_pil_3)
1747
1748
        self.assertEqual(gray_pil_3.mode, 'L', 'mode should be L')
        self.assertEqual(gray_np_3.shape, tuple(x_shape[0:2]), 'should be 1 channel')
1749
1750
1751
1752
1753
1754
        np.testing.assert_equal(gray_np, gray_np_3)

        # Case 4: 1 channel grayscale -> 3 channel grayscale
        trans4 = transforms.Grayscale(num_output_channels=3)
        gray_pil_4 = trans4(x_pil_2)
        gray_np_4 = np.array(gray_pil_4)
1755
1756
        self.assertEqual(gray_pil_4.mode, 'RGB', 'mode should be RGB')
        self.assertEqual(gray_np_4.shape, tuple(x_shape), 'should be 3 channel')
1757
1758
1759
1760
        np.testing.assert_equal(gray_np_4[:, :, 0], gray_np_4[:, :, 1])
        np.testing.assert_equal(gray_np_4[:, :, 1], gray_np_4[:, :, 2])
        np.testing.assert_equal(gray_np, gray_np_4[:, :, 0])

1761
1762
1763
        # Checking if Grayscale can be printed as string
        trans4.__repr__()

1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_grayscale(self):
        """Unit tests for random grayscale transform"""

        # Test Set 1: RGB -> 3 channel grayscale
        random_state = random.getstate()
        random.seed(42)
        x_shape = [2, 2, 3]
        x_np = np.random.randint(0, 256, x_shape, np.uint8)
        x_pil = Image.fromarray(x_np, mode='RGB')
        x_pil_2 = x_pil.convert('L')
        gray_np = np.array(x_pil_2)

        num_samples = 250
        num_gray = 0
        for _ in range(num_samples):
            gray_pil_2 = transforms.RandomGrayscale(p=0.5)(x_pil)
            gray_np_2 = np.array(gray_pil_2)
            if np.array_equal(gray_np_2[:, :, 0], gray_np_2[:, :, 1]) and \
1783
1784
                    np.array_equal(gray_np_2[:, :, 1], gray_np_2[:, :, 2]) and \
                    np.array_equal(gray_np, gray_np_2[:, :, 0]):
1785
1786
1787
1788
                num_gray = num_gray + 1

        p_value = stats.binom_test(num_gray, num_samples, p=0.5)
        random.setstate(random_state)
1789
        self.assertGreater(p_value, 0.0001)
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809

        # Test Set 2: grayscale -> 1 channel grayscale
        random_state = random.getstate()
        random.seed(42)
        x_shape = [2, 2, 3]
        x_np = np.random.randint(0, 256, x_shape, np.uint8)
        x_pil = Image.fromarray(x_np, mode='RGB')
        x_pil_2 = x_pil.convert('L')
        gray_np = np.array(x_pil_2)

        num_samples = 250
        num_gray = 0
        for _ in range(num_samples):
            gray_pil_3 = transforms.RandomGrayscale(p=0.5)(x_pil_2)
            gray_np_3 = np.array(gray_pil_3)
            if np.array_equal(gray_np, gray_np_3):
                num_gray = num_gray + 1

        p_value = stats.binom_test(num_gray, num_samples, p=1.0)  # Note: grayscale is always unchanged
        random.setstate(random_state)
1810
        self.assertGreater(p_value, 0.0001)
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823

        # Test set 3: Explicit tests
        x_shape = [2, 2, 3]
        x_data = [0, 5, 13, 54, 135, 226, 37, 8, 234, 90, 255, 1]
        x_np = np.array(x_data, dtype=np.uint8).reshape(x_shape)
        x_pil = Image.fromarray(x_np, mode='RGB')
        x_pil_2 = x_pil.convert('L')
        gray_np = np.array(x_pil_2)

        # Case 3a: RGB -> 3 channel grayscale (grayscaled)
        trans2 = transforms.RandomGrayscale(p=1.0)
        gray_pil_2 = trans2(x_pil)
        gray_np_2 = np.array(gray_pil_2)
1824
1825
        self.assertEqual(gray_pil_2.mode, 'RGB', 'mode should be RGB')
        self.assertEqual(gray_np_2.shape, tuple(x_shape), 'should be 3 channel')
1826
1827
1828
1829
1830
1831
1832
1833
        np.testing.assert_equal(gray_np_2[:, :, 0], gray_np_2[:, :, 1])
        np.testing.assert_equal(gray_np_2[:, :, 1], gray_np_2[:, :, 2])
        np.testing.assert_equal(gray_np, gray_np_2[:, :, 0])

        # Case 3b: RGB -> 3 channel grayscale (unchanged)
        trans2 = transforms.RandomGrayscale(p=0.0)
        gray_pil_2 = trans2(x_pil)
        gray_np_2 = np.array(gray_pil_2)
1834
1835
        self.assertEqual(gray_pil_2.mode, 'RGB', 'mode should be RGB')
        self.assertEqual(gray_np_2.shape, tuple(x_shape), 'should be 3 channel')
1836
1837
1838
1839
1840
1841
        np.testing.assert_equal(x_np, gray_np_2)

        # Case 3c: 1 channel grayscale -> 1 channel grayscale (grayscaled)
        trans3 = transforms.RandomGrayscale(p=1.0)
        gray_pil_3 = trans3(x_pil_2)
        gray_np_3 = np.array(gray_pil_3)
1842
1843
        self.assertEqual(gray_pil_3.mode, 'L', 'mode should be L')
        self.assertEqual(gray_np_3.shape, tuple(x_shape[0:2]), 'should be 1 channel')
1844
1845
1846
1847
1848
1849
        np.testing.assert_equal(gray_np, gray_np_3)

        # Case 3d: 1 channel grayscale -> 1 channel grayscale (unchanged)
        trans3 = transforms.RandomGrayscale(p=0.0)
        gray_pil_3 = trans3(x_pil_2)
        gray_np_3 = np.array(gray_pil_3)
1850
1851
        self.assertEqual(gray_pil_3.mode, 'L', 'mode should be L')
        self.assertEqual(gray_np_3.shape, tuple(x_shape[0:2]), 'should be 1 channel')
1852
1853
        np.testing.assert_equal(gray_np, gray_np_3)

1854
1855
1856
        # Checking if RandomGrayscale can be printed as string
        trans3.__repr__()

1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
    def test_gaussian_blur_asserts(self):
        np_img = np.ones((100, 100, 3), dtype=np.uint8) * 255
        img = F.to_pil_image(np_img, "RGB")

        with self.assertRaisesRegex(ValueError, r"If kernel_size is a sequence its length should be 2"):
            F.gaussian_blur(img, [3])

        with self.assertRaisesRegex(ValueError, r"If kernel_size is a sequence its length should be 2"):
            F.gaussian_blur(img, [3, 3, 3])
        with self.assertRaisesRegex(ValueError, r"Kernel size should be a tuple/list of two integers"):
            transforms.GaussianBlur([3, 3, 3])

        with self.assertRaisesRegex(ValueError, r"kernel_size should have odd and positive integers"):
            F.gaussian_blur(img, [4, 4])
        with self.assertRaisesRegex(ValueError, r"Kernel size value should be an odd and positive number"):
            transforms.GaussianBlur([4, 4])

        with self.assertRaisesRegex(ValueError, r"kernel_size should have odd and positive integers"):
            F.gaussian_blur(img, [-3, -3])
        with self.assertRaisesRegex(ValueError, r"Kernel size value should be an odd and positive number"):
            transforms.GaussianBlur([-3, -3])

        with self.assertRaisesRegex(ValueError, r"If sigma is a sequence, its length should be 2"):
            F.gaussian_blur(img, 3, [1, 1, 1])
        with self.assertRaisesRegex(ValueError, r"sigma should be a single number or a list/tuple with length 2"):
            transforms.GaussianBlur(3, [1, 1, 1])

        with self.assertRaisesRegex(ValueError, r"sigma should have positive values"):
            F.gaussian_blur(img, 3, -1.0)
        with self.assertRaisesRegex(ValueError, r"If sigma is a single number, it must be positive"):
            transforms.GaussianBlur(3, -1.0)

        with self.assertRaisesRegex(TypeError, r"kernel_size should be int or a sequence of integers"):
            F.gaussian_blur(img, "kernel_size_string")
        with self.assertRaisesRegex(ValueError, r"Kernel size should be a tuple/list of two integers"):
            transforms.GaussianBlur("kernel_size_string")

        with self.assertRaisesRegex(TypeError, r"sigma should be either float or sequence of floats"):
            F.gaussian_blur(img, 3, "sigma_string")
        with self.assertRaisesRegex(ValueError, r"sigma should be a single number or a list/tuple with length 2"):
            transforms.GaussianBlur(3, "sigma_string")

1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
    def _test_randomness(self, fn, trans, configs):
        random_state = random.getstate()
        random.seed(42)
        img = transforms.ToPILImage()(torch.rand(3, 16, 18))

        for p in [0.5, 0.7]:
            for config in configs:
                inv_img = fn(img, **config)

                num_samples = 250
                counts = 0
                for _ in range(num_samples):
                    tranformation = trans(p=p, **config)
                    tranformation.__repr__()
                    out = tranformation(img)
                    if out == inv_img:
                        counts += 1

                p_value = stats.binom_test(counts, num_samples, p=p)
                random.setstate(random_state)
                self.assertGreater(p_value, 0.0001)

    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_invert(self):
        self._test_randomness(
            F.invert,
            transforms.RandomInvert,
            [{}]
        )

    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_posterize(self):
        self._test_randomness(
            F.posterize,
            transforms.RandomPosterize,
            [{"bits": 4}]
        )

    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_solarize(self):
        self._test_randomness(
            F.solarize,
            transforms.RandomSolarize,
            [{"threshold": 192}]
        )

    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_adjust_sharpness(self):
        self._test_randomness(
            F.adjust_sharpness,
            transforms.RandomAdjustSharpness,
            [{"sharpness_factor": 2.0}]
        )

    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_autocontrast(self):
        self._test_randomness(
            F.autocontrast,
            transforms.RandomAutocontrast,
            [{}]
        )

    @unittest.skipIf(stats is None, 'scipy.stats not available')
    def test_random_equalize(self):
        self._test_randomness(
            F.equalize,
            transforms.RandomEqualize,
            [{}]
        )

    def test_autoaugment(self):
        for policy in transforms.AutoAugmentPolicy:
            for fill in [None, 85, (128, 128, 128)]:
                random.seed(42)
                img = Image.open(GRACE_HOPPER)
                transform = transforms.AutoAugment(policy=policy, fill=fill)
                for _ in range(100):
                    img = transform(img)
                transform.__repr__()

1979
    @unittest.skipIf(stats is None, 'scipy.stats not available')
1980
1981
1982
    def test_random_erasing(self):
        img = torch.ones(3, 128, 128)

1983
        t = transforms.RandomErasing(scale=(0.1, 0.1), ratio=(1 / 3, 3.))
1984
1985
        y, x, h, w, v = t.get_params(img, t.scale, t.ratio, [t.value, ])
        aspect_ratio = h / w
1986
1987
1988
        # Add some tolerance due to the rounding and int conversion used in the transform
        tol = 0.05
        self.assertTrue(1 / 3 - tol <= aspect_ratio <= 3 + tol)
1989
1990
1991
1992
1993
1994
1995
1996
1997

        aspect_ratios = []
        random.seed(42)
        trial = 1000
        for _ in range(trial):
            y, x, h, w, v = t.get_params(img, t.scale, t.ratio, [t.value, ])
            aspect_ratios.append(h / w)

        count_bigger_then_ones = len([1 for aspect_ratio in aspect_ratios if aspect_ratio > 1])
1998
1999
        p_value = stats.binom_test(count_bigger_then_ones, trial, p=0.5)
        self.assertGreater(p_value, 0.0001)
2000

2001
2002
2003
        # Checking if RandomErasing can be printed as string
        t.__repr__()

2004

2005
2006
if __name__ == '__main__':
    unittest.main()