Bump ruff from 0.4.4 to 0.4.5 in /requirements (#2080)

* Bump ruff from 0.4.4 to 0.4.5 in /requirements

Bumps [ruff](https://github.com/astral-sh/ruff) from 0.4.4 to 0.4.5.
- [Release notes](https://github.com/astral-sh/ruff/releases)
- [Changelog](https://github.com/astral-sh/ruff/blob/main/CHANGELOG.md)
- [Commits](https://github.com/astral-sh/ruff/compare/v0.4.4...v0.4.5)

---
updated-dependencies:
- dependency-name: ruff
  dependency-type: direct:production
  update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

* min version of numpy now supports type hints

* pre-commit: update ruff

---------

Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Adam J. Stewart <ajstewart426@gmail.com>

.pre-commit-config.yaml

@@ -1,6 +1,6 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.3.7
+    rev: v0.4.5
     hooks:
       - id: ruff
         types_or:

experiments/ssl4eo/landsat/chip_landsat_benchmark.py

@@ -29,7 +29,7 @@ def retrieve_mask_chip(
         mask array
     """
     out_shape = (1, *img_src.shape)
-    mask_chip: 'np.typing.NDArray[np.uint8]' = mask_src.read(
+    mask_chip: np.typing.NDArray[np.uint8] = mask_src.read(
         out_shape=out_shape, window=from_bounds(*img_src.bounds, mask_src.transform)
     )
 

requirements/style.txt

@@ -1,3 +1,3 @@
 # style
 mypy==1.10.0
-ruff==0.4.4
+ruff==0.4.5

tests/datasets/test_utils.py

@@ -561,7 +561,7 @@ def test_nonexisting_directory(tmp_path: Path) -> None:
 
 
 def test_percentile_normalization() -> None:
-    img: 'np.typing.NDArray[np.int_]' = np.array([[1, 2], [98, 100]])
+    img: np.typing.NDArray[np.int_] = np.array([[1, 2], [98, 100]])
 
     img = percentile_normalization(img, 2, 98)
     assert img.min() == 0
@@ -573,7 +573,7 @@ def test_percentile_normalization() -> None:
     [np.uint8, np.uint16, np.uint32, np.int8, np.int16, np.int32, np.int64],
 )
 def test_array_to_tensor(array_dtype: 'np.typing.DTypeLike') -> None:
-    array: 'np.typing.NDArray[Any]' = np.zeros((2,), dtype=array_dtype)
+    array: np.typing.NDArray[Any] = np.zeros((2,), dtype=array_dtype)
     array[0] = np.iinfo(array.dtype).min
     array[1] = np.iinfo(array.dtype).max
     tensor = array_to_tensor(array)

torchgeo/datasets/advance.py

@@ -179,7 +179,7 @@ class ADVANCE(NonGeoDataset):
             the image
         """
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))

torchgeo/datasets/bigearthnet.py

@@ -403,7 +403,7 @@ class BigEarthNet(NonGeoDataset):
                     resampling=Resampling.bilinear,
                 )
                 images.append(array)
-        arrays: 'np.typing.NDArray[np.int_]' = np.stack(images, axis=0)
+        arrays: np.typing.NDArray[np.int_] = np.stack(images, axis=0)
         tensor = torch.from_numpy(arrays).float()
         return tensor
 

torchgeo/datasets/biomassters.py

@@ -196,7 +196,7 @@ class BioMassters(NonGeoDataset):
             target mask
         """
         with rasterio.open(os.path.join(self.root, 'train_agbm', filename), 'r') as src:
-            arr: 'np.typing.NDArray[np.float_]' = src.read()
+            arr: np.typing.NDArray[np.float_] = src.read()
 
         target = torch.from_numpy(arr).float()
         return target

torchgeo/datasets/cloud_cover.py

@@ -183,7 +183,7 @@ class CloudCoverDetection(NonGeoDataset):
             with rasterio.open(path) as image_data:
                 image_array = image_data.read(1).astype(np.int32)
                 images.append(image_array)
-        image_stack: 'np.typing.NDArray[np.int_]' = np.stack(images, axis=0)
+        image_stack: np.typing.NDArray[np.int_] = np.stack(images, axis=0)
         image_tensor = torch.from_numpy(image_stack)
         return image_tensor
 
@@ -200,7 +200,7 @@ class CloudCoverDetection(NonGeoDataset):
         with rasterio.open(label_asset_path) as target_data:
             target_img = target_data.read(1).astype(np.int32)
 
-        target_array: 'np.typing.NDArray[np.int_]' = np.array(target_img)
+        target_array: np.typing.NDArray[np.int_] = np.array(target_img)
         target_tensor = torch.from_numpy(target_array)
         return target_tensor
 

torchgeo/datasets/cowc.py

@@ -145,7 +145,7 @@ class COWC(NonGeoDataset, abc.ABC):
         """
         filename = os.path.join(self.root, self.images[index])
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img)
+            array: np.typing.NDArray[np.int_] = np.array(img)
             tensor = torch.from_numpy(array).float()
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))

torchgeo/datasets/cv4a_kenya_crop_type.py

@@ -235,7 +235,7 @@ class CV4AKenyaCropType(NonGeoDataset):
         )
 
         with Image.open(os.path.join(directory, 'labels.tif')) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img)
+            array: np.typing.NDArray[np.int_] = np.array(img)
             labels = torch.from_numpy(array)
 
         with Image.open(os.path.join(directory, 'field_ids.tif')) as img:
@@ -332,7 +332,7 @@ class CV4AKenyaCropType(NonGeoDataset):
                 f'{band_name}.tif',
             )
             with Image.open(filepath) as band_img:
-                array: 'np.typing.NDArray[np.int_]' = np.array(band_img)
+                array: np.typing.NDArray[np.int_] = np.array(band_img)
                 img[band_index] = torch.from_numpy(array)
 
         return img

torchgeo/datasets/cyclone.py

@@ -159,7 +159,7 @@ class TropicalCyclone(NonGeoDataset):
                 except AttributeError:
                     resample = Image.BILINEAR  # type: ignore[attr-defined]
                 img = img.resize(size=(self.size, self.size), resample=resample)
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array)
             tensor = tensor.permute((2, 0, 1)).float()
             return tensor

torchgeo/datasets/deepglobelandcover.py

@@ -184,7 +184,7 @@ class DeepGlobeLandCover(NonGeoDataset):
         path = self.image_fns[index]
 
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img)
+            array: np.typing.NDArray[np.int_] = np.array(img)
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1)).to(torch.float32)
@@ -201,7 +201,7 @@ class DeepGlobeLandCover(NonGeoDataset):
         """
         path = self.mask_fns[index]
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.uint8]' = np.array(img)
+            array: np.typing.NDArray[np.uint8] = np.array(img)
             array = rgb_to_mask(array, self.colormap)
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW

torchgeo/datasets/dfc2022.py

@@ -235,7 +235,7 @@ class DFC2022(NonGeoDataset):
             the image
         """
         with rasterio.open(path) as f:
-            array: 'np.typing.NDArray[np.float_]' = f.read(
+            array: np.typing.NDArray[np.float_] = f.read(
                 out_shape=shape, out_dtype='float32', resampling=Resampling.bilinear
             )
             tensor = torch.from_numpy(array)
@@ -251,7 +251,7 @@ class DFC2022(NonGeoDataset):
             the target mask
         """
         with rasterio.open(path) as f:
-            array: 'np.typing.NDArray[np.int_]' = f.read(
+            array: np.typing.NDArray[np.int_] = f.read(
                 indexes=1, out_dtype='int32', resampling=Resampling.bilinear
             )
             tensor = torch.from_numpy(array)

torchgeo/datasets/etci2021.py

@@ -204,7 +204,7 @@ class ETCI2021(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array).float()
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))
@@ -221,7 +221,7 @@ class ETCI2021(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('L'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('L'))
             tensor = torch.from_numpy(array)
             tensor = torch.clamp(tensor, min=0, max=1)
             tensor = tensor.to(torch.long)

torchgeo/datasets/eurocrops.py

@@ -246,7 +246,7 @@ class EuroCrops(VectorDataset):
         def apply_cmap(arr: 'np.typing.NDArray[Any]') -> 'np.typing.NDArray[np.float_]':
             # Color 0 as black, while applying default color map for the class indices.
             cmap = plt.get_cmap('viridis')
-            im: 'np.typing.NDArray[np.float_]' = cmap(arr / len(self.class_map))
+            im: np.typing.NDArray[np.float_] = cmap(arr / len(self.class_map))
             im[arr == 0] = 0
             return im
 

torchgeo/datasets/fair1m.py

@@ -308,7 +308,7 @@ class FAIR1M(NonGeoDataset):
             the image
         """
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))

torchgeo/datasets/forestdamage.py

@@ -197,7 +197,7 @@ class ForestDamage(NonGeoDataset):
             the image
         """
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor: Tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))

torchgeo/datasets/geo.py

@@ -904,7 +904,7 @@ class NonGeoClassificationDataset(NonGeoDataset, ImageFolder):  # type: ignore[m
             the image and class label
         """
         img, label = ImageFolder.__getitem__(self, index)
-        array: 'np.typing.NDArray[np.int_]' = np.array(img)
+        array: np.typing.NDArray[np.int_] = np.array(img)
         tensor = torch.from_numpy(array).float()
         # Convert from HxWxC to CxHxW
         tensor = tensor.permute((2, 0, 1))

torchgeo/datasets/gid15.py

@@ -189,7 +189,7 @@ class GID15(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1)).float()
@@ -206,7 +206,7 @@ class GID15(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('L'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('L'))
             tensor = torch.from_numpy(array)
             tensor = tensor.to(torch.long)
             return tensor

torchgeo/datasets/idtrees.py

@@ -265,7 +265,7 @@ class IDTReeS(NonGeoDataset):
         """
         laspy = lazy_import('laspy')
         las = laspy.read(path)
-        array: 'np.typing.NDArray[np.int_]' = np.stack([las.x, las.y, las.z], axis=0)
+        array: np.typing.NDArray[np.int_] = np.stack([las.x, las.y, las.z], axis=0)
         tensor = torch.from_numpy(array)
         return tensor
 
@@ -570,7 +570,7 @@ class IDTReeS(NonGeoDataset):
         path = self.images[index]
         path = path.replace('RGB', 'LAS').replace('.tif', '.las')
         las = laspy.read(path)
-        points: 'np.typing.NDArray[np.int_]' = np.stack(
+        points: np.typing.NDArray[np.int_] = np.stack(
             [las.x, las.y, las.z], axis=0
         ).transpose((1, 0))
         point_cloud = pyvista.PolyData(points)

torchgeo/datasets/landcoverai.py

@@ -360,7 +360,7 @@ class LandCoverAI(LandCoverAIBase, NonGeoDataset):
         """
         filename = os.path.join(self.root, 'output', id_ + '.jpg')
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img)
+            array: np.typing.NDArray[np.int_] = np.array(img)
             tensor = torch.from_numpy(array).float()
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))
@@ -378,7 +378,7 @@ class LandCoverAI(LandCoverAIBase, NonGeoDataset):
         """
         filename = os.path.join(self.root, 'output', id_ + '_m.png')
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('L'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('L'))
             tensor = torch.from_numpy(array).long()
             return tensor
 

torchgeo/datasets/levircd.py

@@ -105,7 +105,7 @@ class LEVIRCDBase(NonGeoDataset, abc.ABC):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array).float()
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))
@@ -122,7 +122,7 @@ class LEVIRCDBase(NonGeoDataset, abc.ABC):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('L'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('L'))
             tensor = torch.from_numpy(array)
             tensor = torch.clamp(tensor, min=0, max=1)
             tensor = tensor.to(torch.long)

torchgeo/datasets/loveda.py

@@ -208,7 +208,7 @@ class LoveDA(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array).float()
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))
@@ -225,7 +225,7 @@ class LoveDA(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('L'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('L'))
             tensor = torch.from_numpy(array)
             tensor = tensor.to(torch.long)
             return tensor

torchgeo/datasets/mapinwild.py

@@ -217,7 +217,7 @@ class MapInWild(NonGeoDataset):
         """
         with rasterio.open(os.path.join(self.root, source, f'{filename}.tif')) as f:
             raw_array = f.read()
-            array: 'np.typing.NDArray[np.int_]' = np.stack(raw_array, axis=0)
+            array: np.typing.NDArray[np.int_] = np.stack(raw_array, axis=0)
             if array.dtype == np.uint16:
                 array = array.astype(np.int32)
             tensor = torch.from_numpy(array).float()

torchgeo/datasets/millionaid.py

@@ -305,7 +305,7 @@ class MillionAID(NonGeoDataset):
             the image
         """
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor: Tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))

torchgeo/datasets/oscd.py

@@ -216,11 +216,11 @@ class OSCD(NonGeoDataset):
         Returns:
             the image
         """
-        images: list['np.typing.NDArray[np.int_]'] = []
+        images: list[np.typing.NDArray[np.int_]] = []
         for path in paths:
             with Image.open(path) as img:
                 images.append(np.array(img))
-        array: 'np.typing.NDArray[np.int_]' = np.stack(images, axis=0).astype(np.int_)
+        array: np.typing.NDArray[np.int_] = np.stack(images, axis=0).astype(np.int_)
         tensor = torch.from_numpy(array).float()
         return tensor
 
@@ -235,7 +235,7 @@ class OSCD(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('L'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('L'))
             tensor = torch.from_numpy(array)
             tensor = torch.clamp(tensor, min=0, max=1)
             tensor = tensor.to(torch.long)
@@ -314,7 +314,7 @@ class OSCD(NonGeoDataset):
             rgb_img = (np.clip((rgb_img - per02) / (per98 - per02), 0, 1) * 255).astype(
                 np.uint8
             )
-            array: 'np.typing.NDArray[np.uint8]' = draw_semantic_segmentation_masks(
+            array: np.typing.NDArray[np.uint8] = draw_semantic_segmentation_masks(
                 torch.from_numpy(rgb_img),
                 sample['mask'],
                 alpha=alpha,

torchgeo/datasets/potsdam.py

@@ -206,7 +206,7 @@ class Potsdam2D(NonGeoDataset):
         """
         path = self.files[index]['mask']
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.uint8]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.uint8] = np.array(img.convert('RGB'))
             array = rgb_to_mask(array, self.colormap)
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW

torchgeo/datasets/reforestree.py

@@ -147,7 +147,7 @@ class ReforesTree(NonGeoDataset):
             the image
         """
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.uint8]' = np.array(img)
+            array: np.typing.NDArray[np.uint8] = np.array(img)
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))

torchgeo/datasets/so2sat.py

@@ -230,7 +230,7 @@ class So2Sat(NonGeoDataset):
         assert split in self.filenames_by_version[version]
 
         self._validate_bands(bands)
-        self.s1_band_indices: 'np.typing.NDArray[np.int_]' = np.array(
+        self.s1_band_indices: np.typing.NDArray[np.int_] = np.array(
             [
                 self.all_s1_band_names.index(b)
                 for b in bands
@@ -240,7 +240,7 @@ class So2Sat(NonGeoDataset):
 
         self.s1_band_names = [self.all_s1_band_names[i] for i in self.s1_band_indices]
 
-        self.s2_band_indices: 'np.typing.NDArray[np.int_]' = np.array(
+        self.s2_band_indices: np.typing.NDArray[np.int_] = np.array(
             [
                 self.all_s2_band_names.index(b)
                 for b in bands

torchgeo/datasets/spacenet.py

@@ -686,7 +686,7 @@ class SpaceNet3(SpaceNet):
         max_speed_bin = 65
         speed_arr_bin = np.arange(min_speed_bin, max_speed_bin + 1)
         bin_size_mph = 10.0
-        speed_cls_arr: 'np.typing.NDArray[np.int_]' = np.array(
+        speed_cls_arr: np.typing.NDArray[np.int_] = np.array(
             [math.ceil(s / bin_size_mph) for s in speed_arr_bin]
         )
 

torchgeo/datasets/usavars.py

@@ -180,7 +180,7 @@ class USAVars(NonGeoDataset):
             the image
         """
         with rasterio.open(path) as f:
-            array: 'np.typing.NDArray[np.int_]' = f.read()
+            array: np.typing.NDArray[np.int_] = f.read()
             tensor = torch.from_numpy(array).float()
             return tensor
 

torchgeo/datasets/vaihingen.py

@@ -190,7 +190,7 @@ class Vaihingen2D(NonGeoDataset):
         """
         path = self.files[index]['image']
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1)).float()
@@ -207,7 +207,7 @@ class Vaihingen2D(NonGeoDataset):
         """
         path = self.files[index]['mask']
         with Image.open(path) as img:
-            array: 'np.typing.NDArray[np.uint8]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.uint8] = np.array(img.convert('RGB'))
             array = rgb_to_mask(array, self.colormap)
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW

torchgeo/datasets/vhr10.py

@@ -285,7 +285,7 @@ class VHR10(NonGeoDataset):
             f'{id_:03d}.jpg',
         )
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img)
+            array: np.typing.NDArray[np.int_] = np.array(img)
             tensor = torch.from_numpy(array)
             tensor = tensor.float()
             # Convert from HxWxC to CxHxW

torchgeo/datasets/xview.py

@@ -163,7 +163,7 @@ class XView2(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('RGB'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('RGB'))
             tensor = torch.from_numpy(array)
             # Convert from HxWxC to CxHxW
             tensor = tensor.permute((2, 0, 1))
@@ -180,7 +180,7 @@ class XView2(NonGeoDataset):
         """
         filename = os.path.join(path)
         with Image.open(filename) as img:
-            array: 'np.typing.NDArray[np.int_]' = np.array(img.convert('L'))
+            array: np.typing.NDArray[np.int_] = np.array(img.convert('L'))
             tensor = torch.from_numpy(array)
             tensor = tensor.to(torch.long)
             return tensor

torchgeo/models/rcf.py

@@ -165,7 +165,7 @@ class RCF(Module):
         sqrt_zca_eigs = np.sqrt(E)
         inv_sqrt_zca_eigs = np.diag(np.power(sqrt_zca_eigs, -1))
         global_ZCA = V.dot(inv_sqrt_zca_eigs).dot(V.T)
-        patches_normalized: 'np.typing.NDArray[np.float32]' = (
+        patches_normalized: np.typing.NDArray[np.float32] = (
             (patches).dot(global_ZCA).dot(global_ZCA.T)
         )