Skip to content

augment_function

data.transform.augmentation.augment_function

ImageType module-attribute 

ImageType = ndarray

MaskType module-attribute 

MaskType = ndarray

OffsetType module-attribute 

OffsetType = tuple[float, float]

INTERPOLATE_METHOD_CV2 module-attribute 

INTERPOLATE_METHOD_CV2 = {'nearest': INTER_NEAREST, 'linear': INTER_LINEAR, 'cubic': INTER_CUBIC}

resize_image 

resize_image(image: Union[Image, Tensor, ndarray], target_size: ImageSizeType, resampling: str = 'bilinear', use_cv2_for_numpy: bool = True) -> Union[Image, Tensor, ndarray]

이미지를 resize합니다.

Parameters:

  • image (Union[Image, Tensor, ndarray]) –

    image data

  • target_size (ImageSizeType) –

    [W, H]

  • resampling (str, default: 'bilinear' ) –

    resampling method. Defaults to "bilinear".

  • use_cv2_for_numpy (bool, default: True ) –

    use cv2 instead of PIL for faster numpy array image resizing. Defaults to True.

Returns:

  • Union[Image, Tensor, ndarray] –

    Union[Image.Image, torch.Tensor, np.ndarray]: resized image

Source code in SaigeToolkit/data/transform/image_function.py 
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
def resize_image(
    image: Union[Image.Image, torch.Tensor, np.ndarray],
    target_size: ImageSizeType,
    resampling: str = "bilinear",
    use_cv2_for_numpy: bool = True,
) -> Union[Image.Image, torch.Tensor, np.ndarray]:
    """이미지를 resize합니다.

    Args:
        image (Union[Image.Image, torch.Tensor, np.ndarray]): image data
        target_size (ImageSizeType): [W, H]
        resampling (str): resampling method. Defaults to "bilinear".
        use_cv2_for_numpy (bool): use cv2 instead of PIL for faster numpy array image resizing. Defaults to True.

    Returns:
        Union[Image.Image, torch.Tensor, np.ndarray]: resized image
    """
    if isinstance(image, Image.Image):
        image = image.resize(target_size, RESAMPLE_PIL[resampling])
    elif isinstance(image, torch.Tensor):  # CHW, BCHW
        is_three_dim = image.ndim == 3
        if is_three_dim:  # CHW -> 1CHW
            image = image.unsqueeze(0)
        image = torch.nn.functional.interpolate(
            image,
            size=(target_size[1], target_size[0]),
            mode=RESAMPLE_TORCH[resampling],
            antialias=None if resampling == "nearest" else True,
        )
        if is_three_dim:
            image = image.squeeze(0)
    elif isinstance(image, np.ndarray):  # HWC, HW
        if use_cv2_for_numpy:
            # NOTE: cv2 resize is different from PIL/torch (inaccurate but fast)
            image = cv2.resize(image, dsize=target_size, interpolation=RESAMPLE_CV2[resampling])
        else:
            is_single_channel = image.ndim == 3 and image.shape[-1] == 1
            if is_single_channel:
                image = image[:, :, 0]
            image = Image.fromarray(image)
            image = image.resize(target_size, RESAMPLE_PIL[resampling])
            image = np.array(image)
            if is_single_channel:
                image = image[:, :, np.newaxis]
    return image

check_value 

check_value(data: Union[int, float], min_value: Union[int, float], max_value: Union[int, float])
Source code in SaigeToolkit/data/transform/function_util.py 
13
14
15
16
17
18
def check_value(data: Union[int, float], min_value: Union[int, float], max_value: Union[int, float]):
    if not (isinstance(data, (int, float))):
        raise AugmentationParameterTypeError

    if not (min_value <= data <= max_value):
        raise AugmentationParameterRangeError

ratio_to_value 

ratio_to_value(ratio: float, min_value: Union[int, float], max_value: Union[int, float])
Source code in SaigeToolkit/data/transform/function_util.py 
36
37
38
39
40
41
def ratio_to_value(ratio: float, min_value: Union[int, float], max_value: Union[int, float]):
    check_value(ratio, 0.0, 1.0)

    min_value, max_value = min(min_value, max_value), max(min_value, max_value)

    return (ratio * (max_value - min_value)) + min_value

support_gray 

support_gray(function)
Source code in SaigeToolkit/data/transform/function_util.py 
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
def support_gray(function):
    @wraps(function)
    def wrapper(image: ImageType, *args, **kwargs):
        is_gray = image.ndim == 2
        if is_gray:
            image = np.stack([image, image, image], axis=-1)

        image = function(image, *args, **kwargs)

        if is_gray:
            image = image[:, :, 0]

        return image

    return wrapper

support_rgba 

support_rgba(function)
Source code in SaigeToolkit/data/transform/function_util.py 
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
def support_rgba(function):
    @wraps(function)
    def wrapper(image: ImageType, *args, **kwargs):
        is_rgba = image.ndim == 3 and image.shape[2] == 4
        if is_rgba:
            mask = image[:, :, 3:4]
            image = image[:, :, :3]

        image = function(image, *args, **kwargs)

        if is_rgba:
            image = np.concatenate([image, mask], axis=2)

        return image

    return wrapper

vertical_flip 

vertical_flip(image: Union[ImageType, MaskType]) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
44
45
def vertical_flip(image: Union[ImageType, MaskType]) -> Union[ImageType, MaskType]:
    return AF.vflip(image)

horizontal_flip 

horizontal_flip(image: Union[ImageType, MaskType]) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
48
49
50
51
52
53
54
def horizontal_flip(image: Union[ImageType, MaskType]) -> Union[ImageType, MaskType]:
    if image.ndim == 3 and image.shape[2] > 1 and image.dtype == np.uint8:
        # Opencv is faster than numpy only in case of
        # non-gray scale 8bits images
        return AF.hflip_cv2(image)

    return AF.hflip(image)

rotate 

rotate(image: Union[ImageType, MaskType], angle: float = 0, interpolation: int = cv2.INTER_LINEAR, border_mode: int = cv2.BORDER_CONSTANT, value: Union[int, float, List[int], List[float]] = 0, crop_border: bool = False) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
def rotate(
    image: Union[ImageType, MaskType],
    angle: float = 0,
    interpolation: int = cv2.INTER_LINEAR,
    border_mode: int = cv2.BORDER_CONSTANT,
    value: Union[int, float, List[int], List[float]] = 0,
    crop_border: bool = False,
) -> Union[ImageType, MaskType]:
    def _rotated_rect_with_max_area(h, w, angle):
        """
        Given a rectangle of size wxh that has been rotated by 'angle' (in
        degrees), computes the width and height of the largest possible
        axis-aligned rectangle (maximal area) within the rotated rectangle.

        Code from: https://stackoverflow.com/questions/16702966/rotate-image-and-crop-out-black-borders
        """

        angle = math.radians(angle)
        width_is_longer = w >= h
        side_long, side_short = (w, h) if width_is_longer else (h, w)

        # since the solutions for angle, -angle and 180-angle are all the same,
        # it is sufficient to look at the first quadrant and the absolute values of sin,cos:
        sin_a, cos_a = abs(math.sin(angle)), abs(math.cos(angle))
        if side_short <= 2.0 * sin_a * cos_a * side_long or abs(sin_a - cos_a) < 1e-10:
            # half constrained case: two crop corners touch the longer side,
            # the other two corners are on the mid-line parallel to the longer line
            x = 0.5 * side_short
            wr, hr = (x / sin_a, x / cos_a) if width_is_longer else (x / cos_a, x / sin_a)
        else:
            # fully constrained case: crop touches all 4 sides
            cos_2a = cos_a * cos_a - sin_a * sin_a
            wr, hr = (w * cos_a - h * sin_a) / cos_2a, (h * cos_a - w * sin_a) / cos_2a

        return dict(
            x_min=max(0, int(w / 2 - wr / 2)),
            x_max=min(w, int(w / 2 + wr / 2)),
            y_min=max(0, int(h / 2 - hr / 2)),
            y_max=min(h, int(h / 2 + hr / 2)),
        )

    check_value(angle, -360.0, 360.0)

    img_out = AFGeometric.rotate(image, angle, interpolation, border_mode, value)
    if crop_border:
        h, w = image.shape[:2]
        crop_bbox_dict = _rotated_rect_with_max_area(h, w, angle)
        x_min = crop_bbox_dict["x_min"]
        y_min = crop_bbox_dict["y_min"]
        x_max = crop_bbox_dict["x_max"]
        y_max = crop_bbox_dict["y_max"]
        img_out = AFCrops.crop(img_out, x_min, y_min, x_max, y_max)
    return img_out

random_rotate90 

random_rotate90(image: Union[ImageType, MaskType], factor: int = 0) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
112
113
114
115
116
117
def random_rotate90(
    image: Union[ImageType, MaskType],
    factor: int = 0,
) -> Union[ImageType, MaskType]:
    check_value(factor, 0, 3)
    return AF.rot90(img=image, factor=factor)

color_jitter 

color_jitter(image: ImageType, brightness: float = 1.0, contrast: float = 1.0, saturation: float = 1.0, hue: float = 0, order: List[int] = [0, 1, 2, 3]) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
@support_rgba
def color_jitter(
    image: ImageType,
    brightness: float = 1.0,
    contrast: float = 1.0,
    saturation: float = 1.0,
    hue: float = 0,
    order: List[int] = [0, 1, 2, 3],
) -> ImageType:
    if not AF.is_rgb_image(image) and not AF.is_grayscale_image(image):
        raise TypeError("ColorJitter transformation expects 1-channel or 3-channel images.")

    check_value(brightness, 0.01, 10.00)
    check_value(contrast, 0.01, 10.00)
    check_value(saturation, 0.01, 10.00)
    check_value(hue, -0.50, 0.50)

    transforms = [
        AF.adjust_brightness_torchvision,
        AF.adjust_contrast_torchvision,
        AF.adjust_saturation_torchvision,
        AF.adjust_hue_torchvision,
    ]
    params = [brightness, contrast, saturation, hue]

    for i in order:
        image = transforms[i](image, params[i])
    return image

blur 

blur(image: ImageType, ksize: int = 3) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
150
151
152
153
def blur(image: ImageType, ksize: int = 3) -> ImageType:
    check_value(ksize, 1, 100)

    return AF.blur(image, ksize)

gaussian_blur 

gaussian_blur(image: ImageType, ksize: int = 3, sigma: float = 0.0) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
156
157
158
159
160
161
162
163
def gaussian_blur(image: ImageType, ksize: int = 3, sigma: float = 0.0) -> ImageType:
    check_value(ksize, 1, 100)
    check_value(sigma, 0.00, 100.00)

    if ksize % 2 != 1:
        ksize = ksize + 1

    return AF.gaussian_blur(image, ksize, sigma)

adjust_gamma 

adjust_gamma(image: ImageType, gamma: float = 0.0) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
166
167
168
169
170
171
def adjust_gamma(image: ImageType, gamma: float = 0.0) -> ImageType:
    check_value(gamma, -1.0, 1.0)

    gamma = gamma + 1

    return AF.gamma_transform(image, gamma=gamma)

adjust_brightness_contrast 

adjust_brightness_contrast(image: ImageType, brightness: float = 0.0, contrast: float = 0.0, brightness_by_max: bool = True) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
174
175
176
177
178
179
180
181
182
183
def adjust_brightness_contrast(
    image: ImageType, brightness: float = 0.0, contrast: float = 0.0, brightness_by_max: bool = True
) -> ImageType:
    check_value(brightness, -1.00, 1.00)
    check_value(contrast, -1.00, 1.00)

    alpha = 1.0 + contrast
    beta = 0.0 + brightness

    return AF.brightness_contrast_adjust(image, alpha, beta, brightness_by_max)

adjust_brightness 

adjust_brightness(image: ImageType, brightness: float = 0.0, brightness_by_max: bool = True) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
186
187
188
189
190
191
192
193
194
def adjust_brightness(
    image: ImageType, brightness: float = 0.0, brightness_by_max: bool = True
) -> ImageType:
    check_value(brightness, -1.00, 1.00)

    alpha = 1.0
    beta = 0.0 + brightness

    return AF.brightness_contrast_adjust(image, alpha, beta, brightness_by_max)

adjust_contrast 

adjust_contrast(image: ImageType, contrast: float = 0.0) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
197
198
199
200
201
202
203
def adjust_contrast(image: ImageType, contrast: float = 0.0) -> ImageType:
    check_value(contrast, -1.00, 1.00)

    alpha = 1.0 + contrast
    beta = 0.0

    return AF.brightness_contrast_adjust(image, alpha, beta)

adjust_hue 

adjust_hue(image: ImageType, hue: float = 0.0) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
206
207
208
209
210
211
212
213
214
215
@support_gray
@support_rgba
def adjust_hue(image: ImageType, hue: float = 0.0) -> ImageType:
    check_value(hue, -1.00, 1.00)

    hue_shift = int(hue * 180)
    sat_shift = 0
    val_shift = 0

    return AF.shift_hsv(image, hue_shift, sat_shift, val_shift)

adjust_saturation 

adjust_saturation(image: ImageType, saturation: float = 0.0) -> ImageType

adjust_saturation

Parameters:

  • image (ImageType) –

    입력 이미지

  • saturation (float, default: 0.0 ) –

    변형 강도, [-1.0, 1.0] 범위, Defaults to 0.0.

Returns:

  • ImageType ( ImageType ) –

    결과 이미지

Note

Ablumentation의 AF.shift_hsv()와 다른 알고리즘을 사용합니다. AF.shift_hsv()의 경우 색이 없는 픽셀을 붉은 색으로 변형합니다. 이 함수의 경우 색이 없는 픽셀은 변형하지 않습니다.

Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
@support_gray
@support_rgba
def adjust_saturation(image: ImageType, saturation: float = 0.0) -> ImageType:
    """adjust_saturation

    Args:
        image (ImageType): 입력 이미지
        saturation (float, optional): 변형 강도, [-1.0, 1.0] 범위, Defaults to 0.0.

    Returns:
        ImageType: 결과 이미지

    Note:
        Ablumentation의 AF.shift_hsv()와 다른 알고리즘을 사용합니다.
        AF.shift_hsv()의 경우 색이 없는 픽셀을 붉은 색으로 변형합니다.
        이 함수의 경우 색이 없는 픽셀은 변형하지 않습니다.

    """
    check_value(saturation, -1.00, 1.00)
    saturation = (saturation + 1) ** 2
    gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
    result = image * saturation + gray[..., None] * (1 - saturation)

    dtype = image.dtype
    if dtype == np.uint8:
        result = np.clip(result, 0, 255)
    elif dtype == np.uint16:
        result = np.clip(result, 0, 65535)
    result = result.astype(dtype)

    return result

iso_noise 

iso_noise(image: ImageType, color_shift: float = 0.05, intensity: float = 0.5, random_state: Optional[int] = None) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
251
252
253
254
255
256
257
258
259
260
261
262
@support_gray
@support_rgba
def iso_noise(
    image: ImageType,
    color_shift: float = 0.05,
    intensity: float = 0.50,
    random_state: Optional[int] = None,
) -> ImageType:
    check_value(color_shift, 0.00, 1.00)
    check_value(intensity, 0.00, 2.00)

    return AF.iso_noise(image, color_shift, intensity, np.random.RandomState(random_state))

image_compression 

image_compression(image: ImageType, quality: int = 100, image_type: str = '.jpeg') -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
265
266
267
268
269
270
271
272
@support_gray
@support_rgba
def image_compression(
    image: ImageType,
    quality: int = 100,
    image_type: str = ".jpeg",
) -> ImageType:
    return AF.image_compression(image, quality, image_type)

sharpen 

sharpen(image: ImageType, sharpening_matrix: ndarray) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
275
276
277
278
279
280
281
@support_gray
@support_rgba
def sharpen(
    image: ImageType,
    sharpening_matrix: np.ndarray,
) -> ImageType:
    return AF.convolve(image, sharpening_matrix)

multiplicative_noise 

multiplicative_noise(image: ImageType, multiplier: ndarray) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
284
285
286
287
288
289
290
@support_gray
@support_rgba
def multiplicative_noise(
    image: ImageType,
    multiplier: np.ndarray,
) -> ImageType:
    return AF.multiply(image, multiplier)

ratio_jitter 

ratio_jitter(image: ImageType, proportion_left: float = 0.0, proportion_right: float = 0.0, proportion_top: float = 0.0, proportion_bottom: float = 0.0, resampling: str = 'bilinear', border_mode: int = cv2.BORDER_CONSTANT, value: Union[int, float, List[int], List[float]] = 0) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
def ratio_jitter(
    image: ImageType,
    proportion_left: float = 0.0,
    proportion_right: float = 0.0,
    proportion_top: float = 0.0,
    proportion_bottom: float = 0.0,
    resampling: str = "bilinear",
    border_mode: int = cv2.BORDER_CONSTANT,
    value: Union[int, float, List[int], List[float]] = 0,
) -> Union[ImageType, MaskType]:
    check_value(proportion_left, -0.50, 0.50)
    check_value(proportion_right, -0.50, 0.50)
    check_value(proportion_top, -0.50, 0.50)
    check_value(proportion_bottom, -0.50, 0.50)

    h_original, w_original = image.shape[:2]

    left = int(w_original * proportion_left)
    right = int(w_original * proportion_right)
    top = int(h_original * proportion_top)
    bottom = int(h_original * proportion_bottom)

    # crop
    crop_left = left if left > 0 else 0
    crop_right = right if right > 0 else 0
    crop_top = top if top > 0 else 0
    crop_bottom = bottom if bottom > 0 else 0

    crop_w = max(w_original - crop_left - crop_right, 1)
    crop_h = max(h_original - crop_top - crop_bottom, 1)

    image = image[crop_top : crop_top + crop_h, crop_left : crop_left + crop_w]

    # padding
    pad_left = 0 if left > 0 else -left
    pad_right = 0 if right > 0 else -right
    pad_top = 0 if top > 0 else -top
    pad_bottom = 0 if bottom > 0 else -bottom

    image = cv2.copyMakeBorder(
        image,
        top=pad_top,
        bottom=pad_bottom,
        left=pad_left,
        right=pad_right,
        borderType=border_mode,
        value=value,
    )

    # resize
    target_size = (w_original, h_original)
    image = resize_image(image, target_size, resampling)

    return image

_zoom_in 

_zoom_in(image: ImageType, ratio: float = 1.0, h_start: float = 0.0, w_start: float = 0.0, resampling: str = 'bilinear') -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
def _zoom_in(
    image: ImageType,
    ratio: float = 1.0,
    h_start: float = 0.0,
    w_start: float = 0.0,
    resampling: str = "bilinear",
) -> Union[ImageType, MaskType]:
    check_value(ratio, 1.00, 100.00)
    check_value(w_start, 0.00, 1.00)
    check_value(h_start, 0.00, 1.00)

    # zoom in/out
    h_original, w_original = image.shape[:2]
    target_size = (int(w_original * ratio), int(h_original * ratio))
    image = resize_image(image, target_size, resampling)

    # crop
    image = AFCrops.random_crop(
        img=image,
        crop_height=h_original,
        crop_width=w_original,
        h_start=min(h_start, 1.0 - 1e-5),
        w_start=min(w_start, 1.0 - 1e-5),
    )

    return image

_zoom_out 

_zoom_out(image: ImageType, ratio: float = 1.0, h_start: float = 0.0, w_start: float = 0.0, resampling: str = 'bilinear', border_mode: int = cv2.BORDER_CONSTANT, value: Union[int, float, List[int], List[float]] = 0) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
def _zoom_out(
    image: ImageType,
    ratio: float = 1.0,
    h_start: float = 0.0,
    w_start: float = 0.0,
    resampling: str = "bilinear",
    border_mode: int = cv2.BORDER_CONSTANT,
    value: Union[int, float, List[int], List[float]] = 0,
) -> Union[ImageType, MaskType]:
    check_value(ratio, 0.01, 1.00)

    # zoom in/out
    h_original, w_original = image.shape[:2]
    target_size = (max(int(w_original * ratio), 1), max(int(h_original * ratio), 1))
    image = resize_image(image, target_size, resampling)

    # pad
    h_current, w_current = image.shape[:2]
    x1, y1, x2, y2 = AFCrops.get_random_crop_coords(
        h_original, w_original, h_current, w_current, h_start, w_start
    )
    pad_top = y1
    pad_bottom = h_original - y2
    pad_left = x1
    pad_right = w_original - x2

    image = cv2.copyMakeBorder(
        image,
        top=pad_top,
        bottom=pad_bottom,
        left=pad_left,
        right=pad_right,
        borderType=border_mode,
        value=value,
    )

    return image

zoom 

zoom(image: ImageType, ratio: float = 1.0, h_start: float = 0.0, w_start: float = 0.0, resampling: str = 'bilinear', border_mode: int = cv2.BORDER_CONSTANT, value: Union[int, float, List[int], List[float]] = 0) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
def zoom(
    image: ImageType,
    ratio: float = 1.0,
    h_start: float = 0.0,
    w_start: float = 0.0,
    resampling: str = "bilinear",
    border_mode: int = cv2.BORDER_CONSTANT,
    value: Union[int, float, List[int], List[float]] = 0,
) -> Union[ImageType, MaskType]:
    check_value(ratio, 0.01, 100.00)
    check_value(h_start, 0.00, 1.00)
    check_value(w_start, 0.00, 1.00)

    if ratio < 1.0:
        output_image = _zoom_out(
            image=image,
            ratio=ratio,
            h_start=h_start,
            w_start=w_start,
            resampling=resampling,
            border_mode=border_mode,
            value=value,
        )
    elif ratio == 1.0:
        output_image = image
    elif ratio > 1.0:
        output_image = _zoom_in(
            image=image,
            ratio=ratio,
            h_start=h_start,
            w_start=w_start,
            resampling=resampling,
        )
    else:
        raise NotImplementedError

    return output_image

random_resized_crop 

random_resized_crop(image: ImageType, h_scale: float = 1.0, w_scale: float = 1.0, h_start: float = 0.0, w_start: float = 0.0, resampling: str = 'bilinear') -> Union[ImageType, MaskType]

Crop the given image to given scale and then resize it to original size.

Parameters:

  • image (ImageType) –

    original image

  • h_scale (float, default: 1.0 ) –

    crop height scale. Defaults to 1.0.

  • w_scale (float, default: 1.0 ) –

    crop width scale. Defaults to 1.0.

  • h_start (float, default: 0.0 ) –

    crop height start ratio. Defaults to 0.0.

  • w_start (float, default: 0.0 ) –

    crop width start ratio. Defaults to 0.0.

  • resampling (str, default: 'bilinear' ) –

    interpolation method. Defaults to "bilinear".

Returns:

Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
def random_resized_crop(
    image: ImageType,
    h_scale: float = 1.0,
    w_scale: float = 1.0,
    h_start: float = 0.0,
    w_start: float = 0.0,
    resampling: str = "bilinear",
) -> Union[ImageType, MaskType]:
    """Crop the given image to given scale and then resize it to original size.

    Args:
        image (ImageType): original image
        h_scale (float, optional): crop height scale. Defaults to 1.0.
        w_scale (float, optional): crop width scale. Defaults to 1.0.
        h_start (float, optional): crop height start ratio. Defaults to 0.0.
        w_start (float, optional): crop width start ratio. Defaults to 0.0.
        resampling (str, optional): interpolation method. Defaults to "bilinear".

    Returns:
        Union[ImageType, MaskType]: augmented data
    """
    check_value(h_scale, 0.01, 1.00)
    check_value(w_scale, 0.01, 1.00)
    check_value(h_start, 0.00, 1.00)
    check_value(w_start, 0.00, 1.00)

    h_original, w_original = image.shape[:2]
    target_height = int(round(h_original * h_scale))
    target_width = int(round(w_original * w_scale))

    # crop
    image = AFCrops.random_crop(
        img=image,
        crop_height=target_height,
        crop_width=target_width,
        h_start=min(h_start, 1.0 - 1e-5),
        w_start=min(w_start, 1.0 - 1e-5),
    )

    # resize
    target_size = (w_original, h_original)
    image = resize_image(image, target_size, resampling)

    return image

random_resized_crop_and_pad 

random_resized_crop_and_pad(image: ImageType, scale: float = 1.0, aspect_ratio: float = 1.0, h_start: float = 0.0, w_start: float = 0.0, height: Optional[int] = None, width: Optional[int] = None, resampling: str = 'bilinear', border_mode: int = cv2.BORDER_CONSTANT, value: Union[int, float, List[int], List[float]] = 0) -> Union[ImageType, MaskType]
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
def random_resized_crop_and_pad(
    image: ImageType,
    scale: float = 1.0,
    aspect_ratio: float = 1.0,
    h_start: float = 0.0,
    w_start: float = 0.0,
    height: Optional[int] = None,
    width: Optional[int] = None,
    resampling: str = "bilinear",
    border_mode: int = cv2.BORDER_CONSTANT,
    value: Union[int, float, List[int], List[float]] = 0,
) -> Union[ImageType, MaskType]:
    check_value(scale, 0.01, 1.00)
    check_value(aspect_ratio, 0.10, 10.00)
    check_value(h_start, 0.00, 1.00)
    check_value(w_start, 0.00, 1.00)

    h_original, w_original = image.shape[:2]
    area = h_original * w_original
    target_area = scale * area

    crop_height = int(round(math.sqrt(target_area / aspect_ratio)))
    crop_width = int(round(math.sqrt(target_area * aspect_ratio)))

    # pad
    pad_top = max(crop_height - h_original, 0)
    pad_bottom = max(crop_height - h_original, 0)
    pad_left = max(crop_width - w_original, 0)
    pad_right = max(crop_width - w_original, 0)

    image = cv2.copyMakeBorder(
        image,
        top=pad_top,
        bottom=pad_bottom,
        left=pad_left,
        right=pad_right,
        borderType=border_mode,
        value=value,
    )

    # crop
    image = AFCrops.random_crop(
        img=image,
        crop_height=crop_height,
        crop_width=crop_width,
        h_start=min(h_start, 1.0 - 1e-5),
        w_start=min(w_start, 1.0 - 1e-5),
    )

    # resize
    target_size = (width or w_original, height or h_original)
    image = resize_image(image, target_size, resampling)

    return image

light_reflect 

light_reflect(image: ImageType, xc: float, yc: float, x_radius: float, y_radius: float, angle: float) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
@support_gray
def light_reflect(
    image: ImageType, xc: float, yc: float, x_radius: float, y_radius: float, angle: float
) -> ImageType:
    check_value(xc, 0.00, 1.00)
    check_value(yc, 0.00, 1.00)
    check_value(x_radius, 0.00, 1.00)
    check_value(y_radius, 0.00, 1.00)
    check_value(angle, 0.00, 360.00)

    if x_radius == 0 or y_radius == 0:
        return image

    image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV)

    h_original, w_original = image.shape[:2]

    xc = ratio_to_value(xc, 0, w_original)
    yc = ratio_to_value(yc, 0, h_original)
    x_radius = ratio_to_value(x_radius, 0, w_original)
    y_radius = ratio_to_value(y_radius, 0, h_original)
    angle = math.radians(angle)

    c, r = np.meshgrid(np.arange(w_original), np.arange(h_original))
    x_rot = (c - xc) * math.cos(angle) + (r - yc) * math.sin(angle)
    y_rot = -(c - xc) * math.sin(angle) + (r - yc) * math.cos(angle)
    d = x_rot * x_rot / x_radius / x_radius + y_rot * y_rot / y_radius / y_radius
    k = 2 - np.power(d, 0.5)
    k[d >= 1] = 1

    image[:, :, 2] = np.clip(image[:, :, 2] * k, 0, 255).astype(np.uint8)

    image = cv2.cvtColor(image, cv2.COLOR_HSV2RGB)

    return image

perspective_transform 

perspective_transform(image: ImageType, offset_top_left: OffsetType, offset_top_right: OffsetType, offset_bottom_right: OffsetType, offset_bottom_left: OffsetType, interpolate_method: str = 'nearest')
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
def perspective_transform(
    image: ImageType,
    offset_top_left: OffsetType,
    offset_top_right: OffsetType,
    offset_bottom_right: OffsetType,
    offset_bottom_left: OffsetType,
    interpolate_method: str = "nearest",
):
    height, width = image.shape[:2]

    for point_offset in [
        offset_top_left,
        offset_top_right,
        offset_bottom_right,
        offset_bottom_left,
    ]:
        offset_x, offset_y = point_offset

        check_value(offset_x, 0, 49)
        check_value(offset_y, 0, 49)

    transform_matrix = calculate_transform_matrix(
        width=width,
        height=height,
        offset_top_left=offset_top_left,
        offset_top_right=offset_top_right,
        offset_bottom_right=offset_bottom_right,
        offset_bottom_left=offset_bottom_left,
    )
    image = cv2.warpPerspective(
        image, transform_matrix, (width, height), flags=INTERPOLATE_METHOD_CV2[interpolate_method]
    )

    return image

calculate_transform_matrix 

calculate_transform_matrix(width: int, height: int, offset_top_left: OffsetType, offset_top_right: OffsetType, offset_bottom_right: OffsetType, offset_bottom_left: OffsetType) -> ndarray

Calculates the perspective transformation matrix using the given width, height, and corner points of a rectangle.

Parameters:

  • width (int) –

    Width of the original image.

  • height (int) –

    Height of the original image.

  • offset_top_left (OffsetType) –

    The offset ratio of the top-left corner point. Represented by the coordinates (x, y) and has a range of [0, 49]. Calculate xand y according to the procedure below.

    x` = width * offset_top_left[0]
y` = height * offset_top_left[1]
  • offset_bottom_right (OffsetType) –

    The offset ratio of the bottom-right corner point. Calculate xand y according to the procedure below.

    x` = width - width * offset_bottom_right[0]
y` = hegiht - height * offset_bottom_right[1]
  • offset_top_right (OffsetType) –

    The offset ratio of the top-left corner point.

  • offset_bottom_left (OffsetType) –

    The offset ratio of the bottom-left corner point.

Returns:

  • ndarray –

    np.ndarray: The 4x3 perspective transform matrix.

Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
def calculate_transform_matrix(
    width: int,
    height: int,
    offset_top_left: OffsetType,
    offset_top_right: OffsetType,
    offset_bottom_right: OffsetType,
    offset_bottom_left: OffsetType,
) -> np.ndarray:
    """
    Calculates the perspective transformation matrix
    using the given width, height, and corner points of a rectangle.

    Parameters:
        width (int): Width of the original image.
        height (int): Height of the original image.
        offset_top_left (OffsetType):
            The offset ratio of the top-left corner point.
            Represented by the coordinates (x, y) and has a range of [0, 49].
            Calculate x` and y` according to the procedure below.

            ```
            x` = width * offset_top_left[0]
            y` = height * offset_top_left[1]
            ```

        offset_bottom_right (OffsetType):
            The offset ratio of the bottom-right corner point.
            Calculate x` and y` according to the procedure below.

            ```
            x` = width - width * offset_bottom_right[0]
            y` = hegiht - height * offset_bottom_right[1]
            ```

        offset_top_right (OffsetType): The offset ratio of the top-left corner point.
        offset_bottom_left (OffsetType): The offset ratio of the bottom-left corner point.

    Returns:
        np.ndarray: The 4x3 perspective transform matrix.

    """
    points_from = np.float32([[0, 0], [width, 0], [width, height], [0, height]])

    point_top_left: OffsetType = (
        (offset_top_left[0] / 100) * width,
        (offset_top_left[1] / 100) * height,
    )
    point_top_right: OffsetType = (
        width - (offset_top_right[0] / 100) * width,
        (offset_top_right[1] / 100) * height,
    )
    point_bottom_right: OffsetType = (
        width - (offset_bottom_right[0] / 100) * width,
        height - (offset_bottom_right[1] / 100) * height,
    )
    point_bottom_left: OffsetType = (
        (offset_bottom_left[0] / 100) * width,
        height - (offset_bottom_left[1] / 100) * height,
    )
    points_to = np.float32(
        [
            point_top_left,
            point_top_right,
            point_bottom_right,
            point_bottom_left,
        ]
    )
    transform_matrix = cv2.getPerspectiveTransform(points_from, points_to)

    return transform_matrix

erase 

erase(image: ImageType, x: int, y: int, w: int, h: int, value: ndarray, inplace: bool = False)
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
@support_gray
@support_rgba
def erase(
    image: ImageType,
    x: int,
    y: int,
    w: int,
    h: int,
    value: np.ndarray,
    inplace: bool = False,
):
    if not inplace:
        image = image.copy()

    image[y : y + h, x : x + w, ...] = value
    return image.astype("uint8")

random_erase 

random_erase(image: ImageType, scale: float = 0.02, aspect_ratio: float = 1.0, value: Union[tuple[int, int, int], None] = None, inplace: bool = False) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
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
@support_gray
@support_rgba
def random_erase(
    image: ImageType,
    scale: float = 0.02,
    aspect_ratio: float = 1.0,
    value: Union[tuple[int, int, int], None] = None,
    inplace: bool = False,
) -> ImageType:

    check_value(scale, 0.0, 0.5)
    check_value(aspect_ratio, 0.2, 5.0)

    if scale == 0.0:
        return image.astype("uint8")

    img_height, img_width, _ = image.shape

    area = img_height * img_width * scale

    _h = math.sqrt(area * aspect_ratio)
    _w = math.sqrt(area / aspect_ratio)

    x = random.randint(0, max(0, img_width - int(_w)))
    y = random.randint(0, max(0, img_height - int(_h)))
    h = min(int(_h), img_height)
    w = min(int(_w), img_width)

    if value is None:
        # black or gray or random
        value_options = {
            "black": (0, 0, 0),
            "gray": (127, 127, 127),
            "random": "random",
        }
        value = value_options[random.choice(list(value_options.keys()))]

    # cast value to script acceptable type
    if value == "random":
        value = (np.random.normal(size=(h, w, 3)) * 255).astype("uint8")
    else:
        value = np.array([float(v) for v in value]).astype("uint8")[None, None, :]

    if not inplace:
        image = image.copy()

    image[y : y + h, x : x + w, ...] = value
    return image.astype("uint8")

grayscale 

grayscale(image: ImageType) -> ndarray
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
770
771
772
773
774
775
776
@support_gray
@support_rgba
def grayscale(
    image: ImageType,
) -> np.ndarray:
    image = AF.to_gray(image)
    return image.astype("uint8")

gauss_noise 

gauss_noise(image: ImageType, intensity: float = 0.0, mean: float = 0.0, per_channel: bool = True)
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
@support_gray
@support_rgba
def gauss_noise(
    image: ImageType,
    intensity: float = 0.0,
    mean: float = 0.0,
    per_channel: bool = True,
):
    var = intensity * 1000
    sigma = var**0.5

    if per_channel:
        gauss = random_utils.normal(mean, sigma, image.shape)
    else:
        gauss = random_utils.normal(mean, sigma, image.shape[:2])
        if len(image.shape) == 3:
            gauss = np.expand_dims(gauss, -1)

    return AF.gauss_noise(image, gauss=gauss)

get_advanced_blur_kernel 

get_advanced_blur_kernel(ksize: int, sigmaX: float, sigmaY: float, angle: float, beta: float, noise_limit: Sequence[float]) -> Any
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
def get_advanced_blur_kernel(
    ksize: int,
    sigmaX: float,
    sigmaY: float,
    angle: float,
    beta: float,
    noise_limit: Sequence[float],
) -> Any:
    # Generate mesh grid centered at zero.
    ax = np.arange(-ksize // 2 + 1.0, ksize // 2 + 1.0)
    # Shape (ksize, ksize, 2)
    grid = np.stack(np.meshgrid(ax, ax), axis=-1)

    # Calculate rotated sigma matrix
    d_matrix = np.array([[sigmaX**2, 0], [0, sigmaY**2]])
    u_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])
    sigma_matrix = np.dot(u_matrix, np.dot(d_matrix, u_matrix.T))

    inverse_sigma = np.linalg.inv(sigma_matrix)
    # Described in "Parameter Estimation For Multivariate Generalized Gaussian Distributions"
    kernel = np.exp(-0.5 * np.power(np.sum(np.dot(grid, inverse_sigma) * grid, 2), beta))
    # Add noise
    noise_matrix = np.random.uniform(*noise_limit, size=[ksize, ksize])
    kernel = kernel * noise_matrix

    # Normalize kernel
    kernel = kernel.astype(np.float32) / np.sum(kernel)

    return kernel

advanced_blur 

advanced_blur(image: ImageType, ksize: int = 1)
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
@support_gray
@support_rgba
def advanced_blur(
    image: ImageType,
    ksize: int = 1,
):
    if ksize % 2 == 0:
        ksize = ksize + 1

    kernel = get_advanced_blur_kernel(
        ksize=ksize,
        sigmaX=0.3 * ((ksize - 1) * 0.5 - 1) + 0.8,
        sigmaY=0.3 * ((ksize - 1) * 0.5 - 1) + 0.8,
        angle=np.deg2rad(np.random.uniform(-180, 180)),
        beta=1.0,
        noise_limit=(0.9, 1.1),
    )
    return AF.convolve(image, kernel=kernel)

elastic_transform 

elastic_transform(image: ImageType, intensity: float = 0.0, alpha: float = 400.0, alpha_affine: float = 0.0, interpolation: int = cv2.INTER_LINEAR, border_mode: int = cv2.BORDER_REFLECT_101, value: Union[int, float, List[int], List[float]] = 0, approximate: bool = False, same_dxdy: bool = False) -> ImageType
Source code in SaigeToolkit/data/transform/augmentation/augment_function.py 
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
def elastic_transform(
    image: ImageType,
    intensity: float = 0.0,
    alpha: float = 400.0,
    alpha_affine: float = 0.0,
    interpolation: int = cv2.INTER_LINEAR,
    border_mode: int = cv2.BORDER_REFLECT_101,
    value: Union[int, float, List[int], List[float]] = 0,
    approximate: bool = False,
    same_dxdy: bool = False,
) -> ImageType:

    sigma = 15 * (1 - intensity)
    return AFGeometric.elastic_transform(
        img=image,
        alpha=alpha,
        sigma=sigma,
        alpha_affine=alpha_affine,
        interpolation=interpolation,
        border_mode=border_mode,
        value=value,
        approximate=approximate,
        same_dxdy=same_dxdy,
    )