Added longest max size augmentation (#44)

* Added longest max size augmentation

* pep fixes

albumentations/augmentations/functional.py

@@ -242,6 +242,17 @@ def blur(img, ksize):
     return cv2.blur(img, (ksize, ksize))
 
 
+def longest_max_size(img, longest_max_size, interpolation):
+    height, width = img.shape[:2]
+
+    scale = longest_max_size / float(max(width, height))
+
+    if scale != 1.0:
+        out_size = tuple(int(dim * scale) for dim in (width, height))
+        img = cv2.resize(img, out_size, interpolation=interpolation)
+    return img
+
+
 def median_blur(img, ksize):
     if img.dtype == np.float32 and ksize not in {3, 5}:
         raise ValueError(

albumentations/augmentations/transforms.py

@@ -12,7 +12,7 @@
            'RandomRotate90', 'Rotate', 'ShiftScaleRotate', 'CenterCrop', 'OpticalDistortion', 'GridDistortion',
            'ElasticTransform', 'HueSaturationValue', 'PadIfNeeded', 'RGBShift', 'RandomBrightness', 'RandomContrast',
            'MotionBlur', 'MedianBlur', 'GaussNoise', 'CLAHE', 'ChannelShuffle', 'InvertImg', 'ToGray',
-           'JpegCompression', 'Cutout', 'ToFloat', 'FromFloat', 'Crop', 'RandomScale']
+           'JpegCompression', 'Cutout', 'ToFloat', 'FromFloat', 'Crop', 'RandomScale', 'LongestMaxSize']
 
 
 class PadIfNeeded(DualTransform):
@@ -37,9 +37,6 @@ def __init__(self, min_height=1024, min_width=1024, p=1.0):
     def apply(self, img, **params):
         return F.pad(img, min_height=self.min_height, min_width=self.min_width)
 
-    def apply_to_bbox(self, bbox, **params):
-        pass
-
 
 class Crop(DualTransform):
     """Crops region from image
@@ -153,6 +150,29 @@ def apply(self, img, **params):
         return F.transpose(img)
 
 
+class LongestMaxSize(DualTransform):
+    """Rescales and image so that maximum side is equal to max_size, keeping th aspect ratio of the initial image.
+
+    Args:
+        p (float): probability of applying the transform. Default: 1.
+        max_size (int): maximum size of the image after the transformation
+
+    Targets:
+        image, mask
+
+    Image types:
+        uint8, float32
+    """
+
+    def __init__(self, max_size=1024, interpolation=cv2.INTER_LINEAR, p=1):
+        super(LongestMaxSize, self).__init__(p)
+        self.interpolation = interpolation
+        self.max_size = max_size
+
+    def apply(self, img, **params):
+        return F.longest_max_size(img)
+
+
 class RandomRotate90(DualTransform):
     """Randomly rotates the input by 90 degrees zero or more times.
 

docs/probabilities.rst

@@ -8,7 +8,7 @@ About probabilities.
 Default probability values
 ******************************
 
-**Compose**, **PadIfNeeded**, **CenterCrop**, **RandomCrop**, **Crop**, **Normalize**, **ToFloat**, **FromFloat**, **ToTensor** have default
+**Compose**, **PadIfNeeded**, **CenterCrop**, **RandomCrop**, **Crop**, **Normalize**, **ToFloat**, **FromFloat**, **ToTensor**, **LongestMaxSize** have default
 probability values equal to **1**. All other are equal to **0.5**
 
 

tests/test_functional.py

@@ -617,6 +617,22 @@ def test_scale(target):
     assert np.array_equal(scaled, expected)
 
 
+@pytest.mark.parametrize('target', ['image', 'mask'])
+def test_longest_max_size(target):
+    img = np.array([
+        [1, 2, 3],
+        [4, 5, 6],
+        [7, 8, 9],
+        [10, 11, 12]], dtype=np.uint8)
+    expected = np.array([[2, 3],
+                         [6, 7],
+                         [10, 11]], dtype=np.uint8)
+
+    img, expected = convert_2d_to_target_format([img, expected], target=target)
+    scaled = F.longest_max_size(img, longest_max_size=3, 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: