Tidy up

src/transformers/models/conditional_detr/image_processing_conditional_detr.py

@@ -1172,7 +1172,7 @@ def preprocess(
                 f" `pathlib.Path` or string object, but is {type(masks_path)} instead."
             )
 
-        # All transformations expect numpy arrays.
+        # All transformations expect numpy arrays
         images = [to_numpy_array(image) for image in images]
 
         # prepare (COCO annotations as a list of Dict -> DETR target as a single Dict per image)

src/transformers/models/deformable_detr/image_processing_deformable_detr.py

@@ -1170,7 +1170,7 @@ def preprocess(
                 f" `pathlib.Path` or string object, but is {type(masks_path)} instead."
             )
 
-        # All transformations expect numpy arrays.
+        # All transformations expect numpy arrays
         images = [to_numpy_array(image) for image in images]
 
         # prepare (COCO annotations as a list of Dict -> DETR target as a single Dict per image)

src/transformers/models/deta/image_processing_deta.py

@@ -834,7 +834,7 @@ def preprocess(
                 f" `pathlib.Path` or string object, but is {type(masks_path)} instead."
             )
 
-        # All transformations expect numpy arrays.
+        # All transformations expect numpy arrays
         images = [to_numpy_array(image) for image in images]
 
         # prepare (COCO annotations as a list of Dict -> DETR target as a single Dict per image)

src/transformers/models/detr/image_processing_detr.py

@@ -1140,7 +1140,7 @@ def preprocess(
                 f" `pathlib.Path` or string object, but is {type(masks_path)} instead."
             )
 
-        # All transformations expect numpy arrays.
+        # All transformations expect numpy arrays
         images = [to_numpy_array(image) for image in images]
 
         # prepare (COCO annotations as a list of Dict -> DETR target as a single Dict per image)

src/transformers/models/flava/image_processing_flava.py

@@ -30,6 +30,7 @@
     ImageInput,
     PILImageResampling,
     make_list_of_images,
+    to_numpy_array,
     valid_images,
 )
 from ...utils import TensorType, is_vision_available, logging
@@ -454,7 +455,8 @@ def _preprocess_image(
         if do_normalize and (image_mean is None or image_std is None):
             raise ValueError("Image mean and std must be specified if do_normalize is True.")
 
-        image = np.array(image)
+        # All transformations expect numpy arrays.
+        image = to_numpy_array(image)
 
         if do_resize:
             image = self.resize(image=image, size=size, resample=resample)

src/transformers/models/glpn/image_processing_glpn.py

@@ -26,6 +26,7 @@
     PILImageResampling,
     get_image_size,
     make_list_of_images,
+    to_numpy_array,
     valid_images,
 )
 from ...utils import TensorType, logging
@@ -174,7 +175,8 @@ def preprocess(
         if not valid_images(images):
             raise ValueError("Invalid image(s)")
 
-        images = [np.array(img) for img in images]
+        # All transformations expect numpy arrays.
+        images = [to_numpy_array(image) for image in images]
 
         if do_resize:
             images = [self.resize(image, size_divisor=size_divisor, resample=resample) for image in images]

src/transformers/models/mask2former/image_processing_mask2former.py

@@ -559,7 +559,7 @@ def _preprocess_image(
         data_format: Optional[Union[str, ChannelDimension]] = None,
     ) -> np.ndarray:
         """Preprocesses a single image."""
-
+        # All transformations expect numpy arrays.
         image = to_numpy_array(image)
         image = self._preprocess(
             image=image,

src/transformers/models/maskformer/image_processing_maskformer.py

@@ -571,8 +571,8 @@ def _preprocess_image(
         data_format: Optional[Union[str, ChannelDimension]] = None,
     ) -> np.ndarray:
         """Preprocesses a single image."""
-
-        image = np.array(image)
+        # All transformations expect numpy arrays.
+        image = to_numpy_array(image)
         image = self._preprocess(
             image=image,
             do_resize=do_resize,
@@ -597,7 +597,7 @@ def _preprocess_mask(
         size_divisor: int = 0,
     ) -> np.ndarray:
         """Preprocesses a single mask."""
-        segmentation_map = np.array(segmentation_map)
+        segmentation_map = to_numpy_array(segmentation_map)
         # Add channel dimension if missing - needed for certain transformations
         added_channel_dim = False
         if segmentation_map.ndim == 2:

src/transformers/models/oneformer/image_processing_oneformer.py

@@ -555,8 +555,8 @@ def _preprocess_image(
         data_format: Optional[Union[str, ChannelDimension]] = None,
     ) -> np.ndarray:
         """Preprocesses a single image."""
-
-        image = np.array(image)
+        # All transformations expect numpy arrays.
+        image = to_numpy_array(image)
         image = self._preprocess(
             image=image,
             do_resize=do_resize,
@@ -579,7 +579,7 @@ def _preprocess_mask(
         size: Dict[str, int] = None,
     ) -> np.ndarray:
         """Preprocesses a single mask."""
-        segmentation_map = np.array(segmentation_map)
+        segmentation_map = to_numpy_array(segmentation_map)
         # Add channel dimension if missing - needed for certain transformations
         added_channel_dim = False
         if segmentation_map.ndim == 2:

src/transformers/models/owlvit/image_processing_owlvit.py

@@ -316,7 +316,7 @@ def preprocess(
                 "torch.Tensor, tf.Tensor or jax.ndarray."
             )
 
-        # All transformations expect numpy arrays.
+        # All transformations expect numpy arrays
         images = [to_numpy_array(image) for image in images]
 
         if do_resize:

src/transformers/models/pix2struct/image_processing_pix2struct.py

@@ -256,7 +256,6 @@ def extract_flattened_patches(self, image: np.ndarray, max_patches: int, patch_s
         _check_torch_version()
 
         # convert to torch
-        image = np.array(image)
         image = to_channel_dimension_format(image, ChannelDimension.FIRST)
         image = torch.from_numpy(image)
 

src/transformers/models/segformer/image_processing_segformer.py

@@ -248,8 +248,8 @@ def _preprocess_image(
         data_format: Optional[Union[str, ChannelDimension]] = None,
     ) -> np.ndarray:
         """Preprocesses a single image."""
-
-        image = np.array(image)
+        # All transformations expect numpy arrays.
+        image = to_numpy_array(image)
         image = self._preprocess(
             image=image,
             do_reduce_labels=False,
@@ -274,7 +274,7 @@ def _preprocess_mask(
         size: Dict[str, int] = None,
     ) -> np.ndarray:
         """Preprocesses a single mask."""
-        segmentation_map = np.array(segmentation_map)
+        segmentation_map = to_numpy_array(segmentation_map)
         # Add channel dimension if missing - needed for certain transformations
         added_channel_dim = False
         if segmentation_map.ndim == 2:

src/transformers/models/tvlt/image_processing_tvlt.py

@@ -33,6 +33,7 @@
     ImageInput,
     PILImageResampling,
     is_valid_image,
+    to_numpy_array,
     valid_images,
 )
 from ...utils import TensorType, logging
@@ -285,7 +286,8 @@ def _preprocess_image(
         if do_normalize and (image_mean is None or image_std is None):
             raise ValueError("Image mean and std must be specified if do_normalize is True.")
 
-        image = np.array(image)
+        # All transformations expect numpy arrays.
+        image = to_numpy_array(image)
 
         if do_resize:
             image = self.resize(image=image, size=size, resample=resample)

src/transformers/models/videomae/image_processing_videomae.py

@@ -34,6 +34,7 @@
     ImageInput,
     PILImageResampling,
     is_valid_image,
+    to_numpy_array,
     valid_images,
 )
 from ...utils import TensorType, is_vision_available, logging
@@ -257,7 +258,8 @@ def _preprocess_image(
         if do_normalize and (image_mean is None or image_std is None):
             raise ValueError("Image mean and std must be specified if do_normalize is True.")
 
-        image = np.array(image)
+        # All transformations expect numpy arrays.
+        image = to_numpy_array(image)
 
         if do_resize:
             image = self.resize(image=image, size=size, resample=resample)

src/transformers/models/vit/image_processing_vit.py

@@ -255,7 +255,7 @@ def preprocess(
         if do_rescale and rescale_factor is None:
             raise ValueError("Rescale factor must be specified if do_rescale is True.")
 
-        # All transformations expect numpy arrays..
+        # All transformations expect numpy arrays.
         images = [to_numpy_array(image) for image in images]
 
         if do_resize:

src/transformers/models/yolos/image_processing_yolos.py

@@ -1073,7 +1073,7 @@ def preprocess(
                 f" `pathlib.Path` or string object, but is {type(masks_path)} instead."
             )
 
-        # All transformations expect numpy arrays.
+        # All transformations expect numpy arrays
         images = [to_numpy_array(image) for image in images]
 
         # prepare (COCO annotations as a list of Dict -> DETR target as a single Dict per image)