Return applied alpha for MixUp (#1572)

* Return applied alpha for MixUp

* Added mix coef to tests

---------

Co-authored-by: Vladimir Iglovikov <iglovikov@gmail.com>
Co-authored-by: Vladimir Iglovikov <ternaus@users.noreply.github.com>

Author: Dipet

.pre-commit-config.yaml

@@ -67,7 +67,7 @@ repos:
       - id: codespell
         additional_dependencies: ["tomli"]
   - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v1.8.0
+    rev: v1.9.0
     hooks:
       - id: mypy
         files: ^albumentations/

albumentations/augmentations/mixing/transforms.py

@@ -37,6 +37,8 @@ class MixUp(ReferenceBasedTransform):
                 - The returned dictionary must include an 'image' key with a numpy array value.
                 - It may also include 'mask', 'global_label' each associated with numpy array values.
             Defaults to a function that assumes input dictionary contains numpy arrays and directly returns it.
+         mix_coef_return_name (str): Name used for the applied alpha coefficient in the returned dictionary.
+            Defaults to "mix_coef".
         alpha (float):
             The alpha parameter for the Beta distribution, influencing the mix's balance. Must be ≥ 0.
             Higher values lead to more uniform mixing. Defaults to 0.4.
@@ -65,10 +67,12 @@ def __init__(
         reference_data: Optional[Union[Generator[ReferenceImage, None, None], Sequence[Any]]] = None,
         read_fn: Callable[[ReferenceImage], Any] = lambda x: {"image": x, "mask": None, "class_label": None},
         alpha: float = 0.4,
+        mix_coef_return_name: str = "mix_coef",
         always_apply: bool = False,
         p: float = 0.5,
     ):
         super().__init__(always_apply, p)
+        self.mix_coef_return_name = mix_coef_return_name
 
         if alpha < 0:
             msg = "Alpha must be >= 0."
@@ -151,3 +155,9 @@ def get_params(self) -> Dict[str, Union[None, float, Dict[str, Any]]]:
         # If mix_data is not None, calculate mix_coef and apply read_fn
         mix_coef = beta(self.alpha, self.alpha)  # Assuming beta is defined elsewhere
         return {"mix_data": self.read_fn(mix_data), "mix_coef": mix_coef}
+
+    def apply_with_params(self, params: Dict[str, Any], *args: Any, **kwargs: Any) -> Dict[str, Any]:
+        res = super().apply_with_params(params, *args, **kwargs)
+        if self.mix_coef_return_name:
+            res[self.mix_coef_return_name] = params["mix_coef"]
+        return res

requirements-dev.txt

@@ -1,5 +1,5 @@
 deepdiff>=6.7.1
-mypy>=1.8.0
+mypy>=1.9.0
 pre_commit>=3.5.0
 pytest>=8.0.2
 pytest_cov>=4.1.0

tests/test_mixing.py

@@ -40,7 +40,7 @@ def complex_read_fn_image(x):
             "reference_data": complex_image_generator(),
             "read_fn": complex_read_fn_image})] )
 def test_image_only(augmentation_cls, params, image):
-    aug = augmentation_cls(p=1, **params)
+    aug = A.Compose([augmentation_cls(p=1, **params)], p=1)
     data = aug(image=image)
     assert data["image"].dtype == np.uint8
 
@@ -58,20 +58,25 @@ def test_image_only(augmentation_cls, params, image):
               ]
 )
 def test_image_global_label(augmentation_cls, params, image, global_label):
-    aug = augmentation_cls(p=1, **params)
+    aug = A.Compose([augmentation_cls(p=1, **params)], p=1)
 
     data = aug(image=image, global_label=global_label)
 
     assert data["image"].dtype == np.uint8
 
-    reference_item = params["read_fn"](aug.reference_data[0])
+    reference_data = params["reference_data"][0]
+
+    reference_item = params["read_fn"](reference_data)
 
     reference_image = reference_item["image"]
     reference_global_label = reference_item["global_label"]
 
+    mix_coef = data["mix_coef"]
+
     mix_coeff_image = find_mix_coef(data["image"], image, reference_image)
     mix_coeff_label = find_mix_coef(data["global_label"], global_label, reference_global_label)
 
+    assert math.isclose(mix_coef, mix_coeff_image, abs_tol=0.01)
     assert math.isclose(mix_coeff_image, mix_coeff_label, abs_tol=0.01)
     assert 0 <= mix_coeff_image <= 1
 
@@ -85,16 +90,19 @@ def test_image_global_label(augmentation_cls, params, image, global_label):
                 "read_fn": lambda x: x})]
 )
 def test_image_mask_global_label(augmentation_cls, params, image, mask, global_label):
-    aug = augmentation_cls(p=1, **params)
+    aug = A.Compose([augmentation_cls(p=1, **params)], p=1)
 
     data = aug(image=image, global_label=global_label, mask=mask)
 
-    assert data["image"].dtype == np.uint8
+    reference_data = params["reference_data"][0]
 
-    mix_coeff_image = find_mix_coef(data["image"], image, aug.reference_data[0]["image"])
-    mix_coeff_mask = find_mix_coef(data["mask"], mask, aug.reference_data[0]["mask"])
-    mix_coeff_label = find_mix_coef(data["global_label"], global_label, aug.reference_data[0]["global_label"])
+    mix_coef = data["mix_coef"]
 
+    mix_coeff_image = find_mix_coef(data["image"], image, reference_data["image"])
+    mix_coeff_mask = find_mix_coef(data["mask"], mask, reference_data["mask"])
+    mix_coeff_label = find_mix_coef(data["global_label"], global_label, reference_data["global_label"])
+
+    assert math.isclose(mix_coef, mix_coeff_image, abs_tol=0.01)
     assert math.isclose(mix_coeff_image, mix_coeff_label, abs_tol=0.01)
     assert math.isclose(mix_coeff_image, mix_coeff_mask, abs_tol=0.01)
     assert 0 <= mix_coeff_image <= 1
@@ -115,6 +123,8 @@ def test_additional_targets(image, mask, global_label):
 
     data = aug(image=image, global_label=global_label, mask=mask, image1=image1, global_label1=global_label1, mask1=mask1)
 
+    mix_coef = data["mix_coef"]
+
     assert data["image"].dtype == np.uint8
 
     mix_coeff_image = find_mix_coef(data["image"], image, reference_data[0]["image"])
@@ -125,6 +135,7 @@ def test_additional_targets(image, mask, global_label):
     mix_coeff_mask1 = find_mix_coef(data["mask1"], mask1, reference_data[0]["mask"])
     mix_coeff_label1 = find_mix_coef(data["global_label1"], global_label1, reference_data[0]["global_label"])
 
+    assert math.isclose(mix_coef, mix_coeff_image, abs_tol=0.01)
     assert math.isclose(mix_coeff_image, mix_coeff_label, abs_tol=0.01)
 
     assert math.isclose(mix_coeff_image, mix_coeff_mask, abs_tol=0.01)
@@ -176,6 +187,9 @@ def test_pipeline(augmentation_cls, params, image, mask, global_label):
 
     assert data["image"].dtype == np.uint8
 
+    mix_coef = data["mix_coef"]
+
     mix_coeff_label = find_mix_coef(data["global_label"], global_label, reference_data[0]["global_label"])
 
+    assert math.isclose(mix_coef, mix_coeff_label, abs_tol=0.01)
     assert 0 <= mix_coeff_label <= 1