Added RandomScale transform (#42)

albumentations/augmentations/functional.py

@@ -69,6 +69,13 @@ def rotate(img, angle, interpolation=cv2.INTER_LINEAR, border_mode=cv2.BORDER_RE
     return img
 
 
+def scale(img, scale, interpolation=cv2.INTER_LINEAR):
+    height, width = img.shape[:2]
+    new_height, new_width = int(height * scale), int(height * scale)
+    img = cv2.resize(img, (new_width, new_height), interpolation=interpolation)
+    return img
+
+
 def shift_scale_rotate(img, angle, scale, dx, dy, interpolation=cv2.INTER_LINEAR, border_mode=cv2.BORDER_REFLECT_101):
     height, width = img.shape[:2]
     center = (width / 2, height / 2)

albumentations/augmentations/transforms.py

@@ -11,7 +11,7 @@
            'RandomRotate90', 'Rotate', 'ShiftScaleRotate', 'CenterCrop', 'OpticalDistortion', 'GridDistortion',
            'ElasticTransform', 'HueSaturationValue', 'PadIfNeeded', 'RGBShift', 'RandomBrightness', 'RandomContrast',
            'MotionBlur', 'MedianBlur', 'GaussNoise', 'CLAHE', 'ChannelShuffle', 'InvertImg', 'ToGray',
-           'JpegCompression', 'Cutout', 'ToFloat', 'FromFloat', 'Crop']
+           'JpegCompression', 'Cutout', 'ToFloat', 'FromFloat', 'Crop', 'RandomScale']
 
 
 class PadIfNeeded(DualTransform):
@@ -212,6 +212,36 @@ def get_params(self):
         return {'angle': np.random.uniform(self.limit[0], self.limit[1])}
 
 
+class RandomScale(DualTransform):
+    """Randomly resizes. Output image size is different from the input image size.
+
+    Args:
+        scale_limit ((float, float) or float): scaling factor range. If scale_limit is a single float value, the
+            range will be (-scale_limit, scale_limit). Default: 0.1.
+        interpolation (OpenCV flag): flag that is used to specify the interpolation algorithm. Should be one of:
+            cv2.INTER_NEAREST, cv2.INTER_LINEAR, cv2.INTER_CUBIC, cv2.INTER_AREA, cv2.INTER_LANCZOS4.
+            Default: cv2.INTER_LINEAR.
+        p (float): probability of applying the transform. Default: 0.5.
+
+    Targets:
+        image, mask
+
+    Image types:
+        uint8, float32
+    """
+
+    def __init__(self, scale_limit=0.1, p=0.5, interpolation=cv2.INTER_LINEAR):
+        super(RandomScale, self).__init__(p)
+        self.scale_limit = to_tuple(scale_limit)
+        self.interpolation = interpolation
+
+    def apply(self, img, scale=0, interpolation=cv2.INTER_LINEAR, **params):
+        return F.scale(img, scale, interpolation)
+
+    def get_params(self):
+        return {'scale': np.random.uniform(1 + self.scale_limit[0], 1 + self.scale_limit[1])}
+
+
 class ShiftScaleRotate(DualTransform):
     """Randomly applies affine transforms: translates, scales and rotates the input.
 

tests/test_functional.py

@@ -596,6 +596,25 @@ def test_from_float_with_max_value_specified(max_value):
     assert_array_almost_equal_nulp(F.from_float(img, dtype=np.uint32, max_value=max_value), expected)
 
 
+@pytest.mark.parametrize('target', ['image', 'mask'])
+def test_scale_rotate(target):
+    img = np.array([
+        [1, 2, 3, 4],
+        [5, 6, 7, 8],
+        [9, 10, 11, 12],
+        [13, 14, 15, 16]], dtype=np.uint8)
+    expected = np.array([[1, 1, 2, 3, 3, 4],
+                         [3, 4, 4, 5, 6, 6],
+                         [5, 6, 7, 7, 8, 8],
+                         [8, 9, 9, 10, 11, 11],
+                         [11, 12, 12, 13, 14, 14],
+                         [13, 13, 14, 15, 15, 16]], dtype=np.uint8)
+
+    img, expected = convert_2d_to_target_format([img, expected], target=target)
+    scaled = F.scale(img, scale=1.5, interpolation=cv2.INTER_LINEAR)
+    assert np.array_equal(scaled, expected)
+
+
 def test_from_float_unknown_dtype():
     img = np.ones((100, 100, 3), dtype=np.float32)
     with pytest.raises(RuntimeError) as exc_info: