Move out input validation into base image processor

src/transformers/image_processing_utils.py

@@ -26,7 +26,7 @@
 from .dynamic_module_utils import custom_object_save
 from .feature_extraction_utils import BatchFeature as BaseBatchFeature
 from .image_transforms import center_crop, normalize, rescale
-from .image_utils import ChannelDimension
+from .image_utils import ChannelDimension, is_scaled_image, to_numpy_array, valid_images
 from .utils import (
     IMAGE_PROCESSOR_NAME,
     PushToHubMixin,
@@ -44,9 +44,60 @@
 if is_vision_available():
     from PIL import Image
 
+    from .image_utils import PILImageResampling
+
 logger = logging.get_logger(__name__)
 
 
+def validate_kwargs(valid_processor_keys: List[str], captured_kwargs: List[str]):
+    unused_keys = set(captured_kwargs).difference(set(valid_processor_keys))
+    if unused_keys:
+        unused_key_str = ", ".join(unused_keys)
+        # TODO raise a warning here instead of simply logging?
+        logger.warning(f"Unused or unrecognized kwargs: {unused_key_str}.")
+
+
+def validate_preprocess_arguments(
+    do_rescale: Optional[bool] = None,
+    rescale_factor: Optional[float] = None,
+    do_normalize: Optional[bool] = None,
+    image_mean: Optional[Union[float, List[float]]] = None,
+    image_std: Optional[Union[float, List[float]]] = None,
+    do_pad: Optional[bool] = None,
+    size_divisibility: Optional[int] = None,
+    do_center_crop: Optional[bool] = None,
+    crop_size: Optional[Dict[str, int]] = None,
+    do_resize: Optional[bool] = None,
+    size: Optional[Dict[str, int]] = None,
+    resample: Optional["PILImageResampling"] = None,
+):
+    """
+    Checks validity of typically used arguments in an `ImageProcessor` `preprocess` method.
+    Raises `ValueError` if arguments incompatibility is caught.
+    Many incompatibilities are model-specific. `do_pad` sometimes needs `size_divisor`,
+    sometimes `size_divisibility`, and sometimes `size`. New models and processors added should follow
+    existing arguments when possible.
+
+    """
+    if do_rescale and rescale_factor is None:
+        raise ValueError("rescale_factor must be specified if do_rescale is True.")
+
+    if do_pad and size_divisibility is None:
+        # Here, size_divisor might be passed as the value of size
+        raise ValueError(
+            "Depending on moel, size_divisibility, size_divisor, pad_size or size must be specified if do_pad is True."
+        )
+
+    if do_normalize and (image_mean is None or image_std is None):
+        raise ValueError("image_mean and image_std must both be specified if do_normalize is True.")
+
+    if do_center_crop and crop_size is None:
+        raise ValueError("crop_size must be specified if do_center_crop is True.")
+
+    if do_resize and (size is None or resample is None):
+        raise ValueError("size and resample must be specified if do_resize is True.")
+
+
 # TODO: Move BatchFeature to be imported by both image_processing_utils and image_processing_utils
 # We override the class string here, but logic is the same.
 class BatchFeature(BaseBatchFeature):
@@ -543,13 +594,64 @@ def fetch_images(self, image_url_or_urls: Union[str, List[str]]):
 
 
 class BaseImageProcessor(ImageProcessingMixin):
+    _valid_processor_keys = None
+
     def __init__(self, **kwargs):
         super().__init__(**kwargs)
 
     def __call__(self, images, **kwargs) -> BatchFeature:
         """Preprocess an image or a batch of images."""
+        self._validate_inputs(images, **kwargs)
         return self.preprocess(images, **kwargs)
 
+    def _validate_preprocess_arguments(self, **kwargs):
+        """Check if the arguments passed to the preprocess method have compatible settings e.g. if `size` is defined when `do_resize` is set to True."""
+        validate_preprocess_arguments(**kwargs)
+
+    def _validate_image_inputs(self, images, segmentation_maps=None, do_rescale=False):
+        """Check if the images and segmentation maps are valid."""
+        if not valid_images(images):
+            raise ValueError(
+                "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, torch.Tensor, tf.Tensor or jax.ndarray."
+            )
+
+        if segmentation_maps is not None and not valid_images(segmentation_maps):
+            raise ValueError(
+                "Invalid segmentation map type. Must be of type PIL.Image.Image, numpy.ndarray, torch.Tensor, tf.Tensor or jax.ndarray."
+            )
+
+        if do_rescale and is_scaled_image(to_numpy_array(images[0])):
+            logger.warning_once(
+                "It looks like you are trying to rescale already rescaled images. If the input"
+                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
+            )
+
+    def _validate_inputs(self, images, segmentation_maps=None, **kwargs):
+        """Check if the arguments passed to the preprocess method are valid."""
+        return_tensors = kwargs.pop("return_tensors", None)
+        input_data_format = kwargs.pop("input_data_format", None)
+        data_format = kwargs.pop("data_format", None)
+
+        if return_tensors not in (None, "np", "pt", "tf", "jax"):
+            raise ValueError("return_tensors should be one of 'np', 'pt', 'tf', 'jax'.")
+
+        if input_data_format not in (None, ChannelDimension.FIRST, ChannelDimension.LAST):
+            raise ValueError("input_data_format should be one of 'channels_first' or 'channels_last'.")
+
+        if data_format not in (None, ChannelDimension.FIRST, ChannelDimension.LAST):
+            raise ValueError("data_format should be one of 'channels_first' or 'channels_last'.")
+
+        if self._valid_processor_keys is None:
+            raise ValueError("Each image processor must define self._valid_processor_keys")
+
+        for key in kwargs:
+            if key not in self._valid_processor_keys:
+                raise ValueError(f"Invalid argument {key} passed to preprocess method.")
+
+        validate_kwargs(captured_kwargs=kwargs.keys(), valid_processor_keys=self._valid_processor_keys)
+        self._validate_preprocess_arguments(**kwargs)
+        self._validate_image_inputs(images, segmentation_maps, do_rescale=kwargs.get("do_rescale", False))
+
     def preprocess(self, images, **kwargs) -> BatchFeature:
         raise NotImplementedError("Each image processor must implement its own preprocess method")
 

src/transformers/image_utils.py

@@ -337,47 +337,6 @@ def load_image(image: Union[str, "PIL.Image.Image"], timeout: Optional[float] =
     return image
 
 
-def validate_preprocess_arguments(
-    do_rescale: Optional[bool] = None,
-    rescale_factor: Optional[float] = None,
-    do_normalize: Optional[bool] = None,
-    image_mean: Optional[Union[float, List[float]]] = None,
-    image_std: Optional[Union[float, List[float]]] = None,
-    do_pad: Optional[bool] = None,
-    size_divisibility: Optional[int] = None,
-    do_center_crop: Optional[bool] = None,
-    crop_size: Optional[Dict[str, int]] = None,
-    do_resize: Optional[bool] = None,
-    size: Optional[Dict[str, int]] = None,
-    resample: Optional["PILImageResampling"] = None,
-):
-    """
-    Checks validity of typically used arguments in an `ImageProcessor` `preprocess` method.
-    Raises `ValueError` if arguments incompatibility is caught.
-    Many incompatibilities are model-specific. `do_pad` sometimes needs `size_divisor`,
-    sometimes `size_divisibility`, and sometimes `size`. New models and processors added should follow
-    existing arguments when possible.
-
-    """
-    if do_rescale and rescale_factor is None:
-        raise ValueError("rescale_factor must be specified if do_rescale is True.")
-
-    if do_pad and size_divisibility is None:
-        # Here, size_divisor might be passed as the value of size
-        raise ValueError(
-            "Depending on moel, size_divisibility, size_divisor, pad_size or size must be specified if do_pad is True."
-        )
-
-    if do_normalize and (image_mean is None or image_std is None):
-        raise ValueError("image_mean and image_std must both be specified if do_normalize is True.")
-
-    if do_center_crop and crop_size is None:
-        raise ValueError("crop_size must be specified if do_center_crop is True.")
-
-    if do_resize and (size is None or resample is None):
-        raise ValueError("size and resample must be specified if do_resize is True.")
-
-
 # In the future we can add a TF implementation here when we have TF models.
 class ImageFeatureExtractionMixin:
     """
@@ -759,11 +718,3 @@ def validate_annotations(
                 "(batch of images) with the following keys: `image_id`, `file_name` and `segments_info`, with "
                 "the latter being a list of annotations in the COCO format."
             )
-
-
-def validate_kwargs(valid_processor_keys: List[str], captured_kwargs: List[str]):
-    unused_keys = set(captured_kwargs).difference(set(valid_processor_keys))
-    if unused_keys:
-        unused_key_str = ", ".join(unused_keys)
-        # TODO raise a warning here instead of simply logging?
-        logger.warning(f"Unused or unrecognized kwargs: {unused_key_str}.")

src/transformers/models/beit/image_processing_beit.py

@@ -28,12 +28,8 @@
     ImageInput,
     PILImageResampling,
     infer_channel_dimension_format,
-    is_scaled_image,
     make_list_of_images,
     to_numpy_array,
-    valid_images,
-    validate_kwargs,
-    validate_preprocess_arguments,
 )
 from ...utils import TensorType, is_torch_available, is_torch_tensor, is_vision_available, logging
 
@@ -257,11 +253,6 @@ def _preprocess_image(
         """Preprocesses a single image."""
         # All transformations expect numpy arrays.
         image = to_numpy_array(image)
-        if is_scaled_image(image) and do_rescale:
-            logger.warning_once(
-                "It looks like you are trying to rescale already rescaled images. If the input"
-                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
-            )
         if input_data_format is None:
             input_data_format = infer_channel_dimension_format(image)
         image = self._preprocess(
@@ -418,37 +409,11 @@ def preprocess(
         image_std = image_std if image_std is not None else self.image_std
         do_reduce_labels = do_reduce_labels if do_reduce_labels is not None else self.do_reduce_labels
 
-        validate_kwargs(captured_kwargs=kwargs.keys(), valid_processor_keys=self._valid_processor_keys)
-
         images = make_list_of_images(images)
 
         if segmentation_maps is not None:
             segmentation_maps = make_list_of_images(segmentation_maps, expected_ndims=2)
 
-        if segmentation_maps is not None and not valid_images(segmentation_maps):
-            raise ValueError(
-                "Invalid segmentation_maps type. Must be of type PIL.Image.Image, numpy.ndarray, "
-                "torch.Tensor, tf.Tensor or jax.ndarray."
-            )
-        if not valid_images(images):
-            raise ValueError(
-                "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
-                "torch.Tensor, tf.Tensor or jax.ndarray."
-            )
-
-        validate_preprocess_arguments(
-            do_rescale=do_rescale,
-            rescale_factor=rescale_factor,
-            do_normalize=do_normalize,
-            image_mean=image_mean,
-            image_std=image_std,
-            do_center_crop=do_center_crop,
-            crop_size=crop_size,
-            do_resize=do_resize,
-            size=size,
-            resample=resample,
-        )
-
         images = [
             self._preprocess_image(
                 image=img,

src/transformers/models/bit/image_processing_bit.py

@@ -32,12 +32,8 @@
     ImageInput,
     PILImageResampling,
     infer_channel_dimension_format,
-    is_scaled_image,
     make_list_of_images,
     to_numpy_array,
-    valid_images,
-    validate_kwargs,
-    validate_preprocess_arguments,
 )
 from ...utils import TensorType, is_vision_available, logging
 
@@ -274,42 +270,15 @@ def preprocess(
         image_std = image_std if image_std is not None else self.image_std
         do_convert_rgb = do_convert_rgb if do_convert_rgb is not None else self.do_convert_rgb
 
-        validate_kwargs(captured_kwargs=kwargs.keys(), valid_processor_keys=self._valid_processor_keys)
-
         images = make_list_of_images(images)
 
-        if not valid_images(images):
-            raise ValueError(
-                "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
-                "torch.Tensor, tf.Tensor or jax.ndarray."
-            )
-
-        validate_preprocess_arguments(
-            do_rescale=do_rescale,
-            rescale_factor=rescale_factor,
-            do_normalize=do_normalize,
-            image_mean=image_mean,
-            image_std=image_std,
-            do_center_crop=do_center_crop,
-            crop_size=crop_size,
-            do_resize=do_resize,
-            size=size,
-            resample=resample,
-        )
-
         # PIL RGBA images are converted to RGB
         if do_convert_rgb:
             images = [convert_to_rgb(image) for image in images]
 
         # All transformations expect numpy arrays.
         images = [to_numpy_array(image) for image in images]
 
-        if is_scaled_image(images[0]) and do_rescale:
-            logger.warning_once(
-                "It looks like you are trying to rescale already rescaled images. If the input"
-                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
-            )
-
         if input_data_format is None:
             # We assume that all images have the same channel dimension format.
             input_data_format = infer_channel_dimension_format(images[0])

src/transformers/models/blip/image_processing_blip.py

@@ -27,12 +27,8 @@
     ImageInput,
     PILImageResampling,
     infer_channel_dimension_format,
-    is_scaled_image,
     make_list_of_images,
     to_numpy_array,
-    valid_images,
-    validate_kwargs,
-    validate_preprocess_arguments,
 )
 from ...utils import TensorType, is_vision_available, logging
 
@@ -250,37 +246,13 @@ def preprocess(
 
         images = make_list_of_images(images)
 
-        validate_kwargs(captured_kwargs=kwargs.keys(), valid_processor_keys=self._valid_processor_keys)
-
-        if not valid_images(images):
-            raise ValueError(
-                "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
-                "torch.Tensor, tf.Tensor or jax.ndarray."
-            )
-
-        validate_preprocess_arguments(
-            do_rescale=do_rescale,
-            rescale_factor=rescale_factor,
-            do_normalize=do_normalize,
-            image_mean=image_mean,
-            image_std=image_std,
-            do_resize=do_resize,
-            size=size,
-            resample=resample,
-        )
         # PIL RGBA images are converted to RGB
         if do_convert_rgb:
             images = [convert_to_rgb(image) for image in images]
 
         # All transformations expect numpy arrays.
         images = [to_numpy_array(image) for image in images]
 
-        if is_scaled_image(images[0]) and do_rescale:
-            logger.warning_once(
-                "It looks like you are trying to rescale already rescaled images. If the input"
-                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
-            )
-
         if input_data_format is None:
             # We assume that all images have the same channel dimension format.
             input_data_format = infer_channel_dimension_format(images[0])