Image augmentation

autoPyTorch/pipeline/components/setup/augmentation/image/ImageAugmenter.py

@@ -39,10 +39,6 @@ def get_components() -> Dict[str, BaseImageAugmenter]:
         Dict[str, BaseImageAugmenter]: all BaseImageAugmenter components available
             as choices
     """
-    # Error in implementation of CropToFixedSize augmenter in imgaug
-    # It returns a list of arrays instead of an array
-    # if 'ZeroPadAndCrop' in _augmenters:
-    #     del _augmenters['ZeroPadAndCrop']
     components = OrderedDict()
     components.update(_augmenters)
     components.update(_addons.components)

autoPyTorch/pipeline/components/setup/augmentation/image/RandomAffine.py

@@ -17,13 +17,13 @@
 
 
 class RandomAffine(BaseImageAugmenter):
-    def __init__(self, use_augmenter: bool = True, scale_min: float = 0, scale_offset: float = 0.2,
-                 translate_percent_min: float = 0, translate_percent_offset: float = 0.3,
-                 shear: int = 30, rotate: int = 45, random_state: Optional[Union[int, np.random.RandomState]] = None):
+    def __init__(self, use_augmenter: bool = True, scale_offset: float = 0.2,
+                 translate_percent_offset: float = 0.3, shear: int = 30,
+                 rotate: int = 45, random_state: Optional[Union[int, np.random.RandomState]] = None):
         super().__init__(use_augmenter=use_augmenter)
         self.random_state = random_state
-        self.scale = (scale_min, scale_min + scale_offset)
-        self.translate_percent = (translate_percent_min, translate_percent_min + translate_percent_offset)
+        self.scale = (1, 1 - scale_offset)
+        self.translate_percent = (0, translate_percent_offset)
         self.shear = (-shear, shear)
         self.rotate = (-rotate, rotate)
 
@@ -41,24 +41,20 @@ def get_hyperparameter_search_space(
     ) -> ConfigurationSpace:
 
         cs = ConfigurationSpace()
-        scale_min = UniformFloatHyperparameter('scale_min', lower=0, upper=0.99, default_value=0)
-        scale_offset = UniformFloatHyperparameter('scale_offset', lower=0, upper=0.99, default_value=0.2)
-        translate_percent_min = UniformFloatHyperparameter('translate_percent_min', lower=0, upper=0.99,
-                                                           default_value=0)
-        translate_percent_offset = UniformFloatHyperparameter('translate_percent_offset', lower=0, upper=0.99,
-                                                              default_value=0.3)
+        scale_offset = UniformFloatHyperparameter('scale_offset', lower=0, upper=0.4, default_value=0.2)
+
+        translate_percent_offset = UniformFloatHyperparameter('translate_percent_offset', lower=0, upper=0.4,
+                                                              default_value=0.2)
         shear = UniformIntegerHyperparameter('shear', lower=0, upper=45, default_value=30)
         rotate = UniformIntegerHyperparameter('rotate', lower=0, upper=360, default_value=45)
 
         use_augmenter = CategoricalHyperparameter('use_augmenter', choices=[True, False])
 
-        cs.add_hyperparameters([scale_min, scale_offset, translate_percent_min, translate_percent_offset])
+        cs.add_hyperparameters([scale_offset, translate_percent_offset])
         cs.add_hyperparameters([shear, rotate, use_augmenter])
 
         # only add hyperparameters to configuration space if we are using the augmenter
-        cs.add_condition(CS.EqualsCondition(scale_min, use_augmenter, True))
         cs.add_condition(CS.EqualsCondition(scale_offset, use_augmenter, True))
-        cs.add_condition(CS.EqualsCondition(translate_percent_min, use_augmenter, True))
         cs.add_condition(CS.EqualsCondition(translate_percent_offset, use_augmenter, True))
         cs.add_condition(CS.EqualsCondition(shear, use_augmenter, True))
         cs.add_condition(CS.EqualsCondition(rotate, use_augmenter, True))

autoPyTorch/pipeline/components/setup/augmentation/image/Resize.py

@@ -1,6 +1,5 @@
 from typing import Any, Dict, Optional, Union
 
-import ConfigSpace as CS
 from ConfigSpace.configuration_space import ConfigurationSpace
 from ConfigSpace.hyperparameters import (
     CategoricalHyperparameter,
@@ -15,17 +14,16 @@
 
 
 class Resize(BaseImageAugmenter):
-    def __init__(self, interpolation: str = 'linear', use_augmenter: bool = True,
+    def __init__(self, use_augmenter: bool = True,
                  random_state: Optional[Union[int, np.random.RandomState]] = None):
         super().__init__(use_augmenter=use_augmenter)
         self.random_state = random_state
-        self.interpolation = interpolation
 
     def fit(self, X: Dict[str, Any], y: Any = None) -> BaseImageAugmenter:
         self.check_requirements(X, y)
         if self.use_augmenter:
             self.augmenter: Augmenter = iaa.Resize(size=(X['image_height'], X['image_width']),
-                                                   interpolation=self.interpolation, name=self.get_properties()['name'])
+                                                   interpolation='linear', name=self.get_properties()['name'])
 
         return self
 
@@ -53,13 +51,8 @@ def get_hyperparameter_search_space(
     ) -> ConfigurationSpace:
 
         cs = ConfigurationSpace()
-        interpolation = CategoricalHyperparameter('interpolation', choices=['nearest', 'linear', 'area', 'cubic'],
-                                                  default_value='linear')
         use_augmenter = CategoricalHyperparameter('use_augmenter', choices=[True, False])
-        cs.add_hyperparameters([interpolation, use_augmenter])
-
-        # only add hyperparameters to configuration space if we are using the augmenter
-        cs.add_condition(CS.EqualsCondition(interpolation, use_augmenter, True))
+        cs.add_hyperparameters([use_augmenter])
 
         return cs
 

autoPyTorch/pipeline/components/setup/augmentation/image/ZeroPadAndCrop.py

@@ -14,20 +14,18 @@
 
 
 class ZeroPadAndCrop(BaseImageAugmenter):
-    def __init__(self, x_position: float = 0, y_position: float = 0,
-                 percent: float = 0.1, random_state: Optional[Union[int, np.random.RandomState]] = None):
+    def __init__(self, percent: float = 0.1,
+                 random_state: Optional[Union[int, np.random.RandomState]] = None):
         super().__init__()
         self.random_state = random_state
         self.percent = percent
-        self.position = (x_position, y_position)
         self.pad_augmenter: Optional[Augmenter] = None
         self.crop_augmenter: Optional[Augmenter] = None
 
     def fit(self, X: Dict[str, Any], y: Any = None) -> BaseImageAugmenter:
         self.check_requirements(X, y)
         self.pad_augmenter = iaa.Pad(percent=self.percent, keep_size=False)
-        self.crop_augmenter = iaa.CropToFixedSize(height=X['image_height'], width=X['image_width'],
-                                                  position=self.position)
+        self.crop_augmenter = iaa.CropToFixedSize(height=X['image_height'], width=X['image_width'])
         self.augmenter: Augmenter = iaa.Sequential([
             self.pad_augmenter,
             self.crop_augmenter
@@ -59,10 +57,8 @@ def get_hyperparameter_search_space(
     ) -> ConfigurationSpace:
 
         cs = ConfigurationSpace()
-        x_position = UniformFloatHyperparameter('x_position', lower=0, upper=1, default_value=0)
-        y_position = UniformFloatHyperparameter('y_position', lower=0, upper=1, default_value=0)
         percent = UniformFloatHyperparameter('percent', lower=0, upper=0.5, default_value=0.1)
-        cs.add_hyperparameters([x_position, y_position, percent])
+        cs.add_hyperparameters([percent])
         return cs
 
     @staticmethod

test/test_pipeline/components/test_setup_image_augmenter.py

@@ -10,12 +10,10 @@
 class TestImageAugmenter(unittest.TestCase):
     def test_every_augmenter(self):
         image_augmenter = ImageAugmenter()
-        configuration = image_augmenter.get_hyperparameter_search_space().sample_configuration()
+        configuration = image_augmenter.get_hyperparameter_search_space().get_default_configuration()
         image_augmenter = image_augmenter.set_hyperparameters(configuration=configuration)
         X = dict(X_train=np.random.randint(0, 255, (8, 3, 16, 16), dtype=np.uint8), image_height=16, image_width=16)
         for name, augmenter in image_augmenter.available_augmenters.items():
-            if not augmenter.use_augmenter:
-                continue
             augmenter = augmenter.fit(X)
             # check if augmenter in the component has correct name
             self.assertEqual(augmenter.get_image_augmenter().name, name)