SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Removal of deprecated vision methods and specify deprecation versions (#24570)

Browse Source 
* Removal of deprecated methods and specify versions

* Fix tests
This commit is contained in:
amyeroberts
2023-06-29 15:09:51 +01:00
committed by GitHub
parent 77db28dc52
commit b324557aac
32 changed files with 144 additions and 760 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
docs/source/en/model_doc/conditional_detr.md
View File

@@ -43,7 +43,6 @@ This model was contributed by [DepuMeng](https://huggingface.co/DepuMeng). The o
[[autodoc]] ConditionalDetrImageProcessor
- preprocess
- pad_and_create_pixel_mask
- post_process_object_detection
- post_process_instance_segmentation
- post_process_semantic_segmentation
@@ -53,7 +52,6 @@ This model was contributed by [DepuMeng](https://huggingface.co/DepuMeng). The o
[[autodoc]] ConditionalDetrFeatureExtractor
- __call__
- pad_and_create_pixel_mask
- post_process_object_detection
- post_process_instance_segmentation
- post_process_semantic_segmentation

2
docs/source/en/model_doc/deformable_detr.md
View File

@@ -52,14 +52,12 @@ If you're interested in submitting a resource to be included here, please feel f
[[autodoc]] DeformableDetrImageProcessor
- preprocess
- pad_and_create_pixel_mask
- post_process_object_detection
## DeformableDetrFeatureExtractor
[[autodoc]] DeformableDetrFeatureExtractor
- __call__
- pad_and_create_pixel_mask
- post_process_object_detection
## DeformableDetrConfig

1
docs/source/en/model_doc/detr.md
View File

@@ -190,7 +190,6 @@ If you're interested in submitting a resource to be included here, please feel f
[[autodoc]] DetrFeatureExtractor
- __call__
- pad_and_create_pixel_mask
- post_process_object_detection
- post_process_semantic_segmentation
- post_process_instance_segmentation

4
docs/source/en/preprocessing.md
View File

@@ -449,13 +449,13 @@ or segmentation maps.
### Pad
In some cases, for instance, when fine-tuning [DETR](./model_doc/detr), the model applies scale augmentation at training
time. This may cause images to be different sizes in a batch. You can use [`DetrImageProcessor.pad_and_create_pixel_mask`]
time. This may cause images to be different sizes in a batch. You can use [`DetrImageProcessor.pad`]
from [`DetrImageProcessor`] and define a custom `collate_fn` to batch images together.
```py
>>> def collate_fn(batch):
... pixel_values = [item["pixel_values"] for item in batch]
... encoding = image_processor.pad_and_create_pixel_mask(pixel_values, return_tensors="pt")
... encoding = image_processor.pad(pixel_values, return_tensors="pt")
... labels = [item["labels"] for item in batch]
... batch = {}
... batch["pixel_values"] = encoding["pixel_values"]

2
docs/source/en/tasks/object_detection.md
View File

@@ -305,7 +305,7 @@ to indicate which pixels are real (1) and which are padding (0).
```py
>>> def collate_fn(batch):
... pixel_values = [item["pixel_values"] for item in batch]
... encoding = image_processor.pad_and_create_pixel_mask(pixel_values, return_tensors="pt")
... encoding = image_processor.pad(pixel_values, return_tensors="pt")
... labels = [item["labels"] for item in batch]
... batch = {}
... batch["pixel_values"] = encoding["pixel_values"]

4
docs/source/ko/preprocessing.md
View File

@@ -456,12 +456,12 @@ array([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
예를 들어, [DETR](./model_doc/detr)와 같은 경우에는 모델이 훈련할 때 크기 조정 증강을 적용합니다.
이로 인해 배치 내 이미지 크기가 달라질 수 있습니다.
[`DetrImageProcessor`]의 [`DetrImageProcessor.pad_and_create_pixel_mask`]를 사용하고 사용자 정의 `collate_fn`을 정의해서 배치 이미지를 처리할 수 있습니다.
[`DetrImageProcessor`]의 [`DetrImageProcessor.pad`]를 사용하고 사용자 정의 `collate_fn`을 정의해서 배치 이미지를 처리할 수 있습니다.
```py
>>> def collate_fn(batch):
... pixel_values = [item["pixel_values"] for item in batch]
... encoding = image_processor.pad_and_create_pixel_mask(pixel_values, return_tensors="pt")
... encoding = image_processor.pad(pixel_values, return_tensors="pt")
... labels = [item["labels"] for item in batch]
... batch = {}
... batch["pixel_values"] = encoding["pixel_values"]

2
docs/source/ko/tasks/object_detection.md
View File

@@ -297,7 +297,7 @@ DatasetDict({
```py
>>> def collate_fn(batch):
... pixel_values = [item["pixel_values"] for item in batch]
... encoding = image_processor.pad_and_create_pixel_mask(pixel_values, return_tensors="pt")
... encoding = image_processor.pad(pixel_values, return_tensors="pt")
... labels = [item["labels"] for item in batch]
... batch = {}
... batch["pixel_values"] = encoding["pixel_values"]

26
src/transformers/image_transforms.py
View File

@@ -24,7 +24,6 @@ from .image_utils import (
get_channel_dimension_axis,
get_image_size,
infer_channel_dimension_format,
to_numpy_array,
)
from .utils import ExplicitEnum, TensorType, is_jax_tensor, is_tf_tensor, is_torch_tensor
from .utils.import_utils import (
@@ -345,18 +344,6 @@ def normalize(
data_format (`ChannelDimension`, *optional*):
The channel dimension format of the output image. If unset, will use the inferred format from the input.
"""
requires_backends(normalize, ["vision"])
if isinstance(image, PIL.Image.Image):
warnings.warn(
"PIL.Image.Image inputs are deprecated and will be removed in v4.26.0. Please use numpy arrays instead.",
FutureWarning,
)
# Convert PIL image to numpy array with the same logic as in the previous feature extractor normalize -
# casting to numpy array and dividing by 255.
image = to_numpy_array(image)
image = rescale(image, scale=1 / 255)
if not isinstance(image, np.ndarray):
raise ValueError("image must be a numpy array")
@@ -418,15 +405,10 @@ def center_crop(
"""
requires_backends(center_crop, ["vision"])
if isinstance(image, PIL.Image.Image):
warnings.warn(
"PIL.Image.Image inputs are deprecated and will be removed in v4.26.0. Please use numpy arrays instead.",
FutureWarning,
)
image = to_numpy_array(image)
return_numpy = False if return_numpy is None else return_numpy
else:
return_numpy = True if return_numpy is None else return_numpy
if return_numpy is not None:
warnings.warn("return_numpy is deprecated and will be removed in v.4.33", FutureWarning)
return_numpy = True if return_numpy is None else return_numpy
if not isinstance(image, np.ndarray):
raise ValueError(f"Input image must be of type np.ndarray, got {type(image)}")

9
src/transformers/models/beit/image_processing_beit.py
View File

@@ -128,15 +128,6 @@ class BeitImageProcessor(BaseImageProcessor):
self.image_std = image_std if image_std is not None else IMAGENET_STANDARD_STD
self.do_reduce_labels = do_reduce_labels
@property
def reduce_labels(self) -> bool:
warnings.warn(
"The `reduce_labels` property is deprecated and will be removed in v4.27. Please use"
" `do_reduce_labels` instead.",
FutureWarning,
)
return self.do_reduce_labels
@classmethod
def from_dict(cls, image_processor_dict: Dict[str, Any], **kwargs):
"""

37
src/transformers/models/bridgetower/image_processing_bridgetower.py
View File

@@ -14,7 +14,6 @@
# limitations under the License.
"""Image processor class for BridgeTower."""
import warnings
from typing import Any, Dict, Iterable, List, Optional, Tuple, Union
import numpy as np
@@ -352,42 +351,6 @@ class BridgeTowerImageProcessor(BaseImageProcessor):
return BatchFeature(data=data, tensor_type=return_tensors)
# Copied from transformers.models.vilt.image_processing_vilt.ViltImageProcessor.pad_and_create_pixel_mask
def pad_and_create_pixel_mask(
self,
pixel_values_list: List[ImageInput],
return_tensors: Optional[Union[str, TensorType]] = None,
data_format: Optional[ChannelDimension] = None,
) -> BatchFeature:
"""
Pads a batch of images with zeros to the size of largest height and width in the batch and returns their
corresponding pixel mask.
Args:
images (`List[np.ndarray]`):
Batch of images to pad.
return_tensors (`str` or `TensorType`, *optional*):
The type of tensors to return. Can be one of:
- Unset: Return a list of `np.ndarray`.
- `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
- `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
- `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
- `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
data_format (`str` or `ChannelDimension`, *optional*):
The channel dimension format of the image. If not provided, it will be the same as the input image.
"""
warnings.warn(
"This method is deprecated and will be removed in v4.26.0. Please use pad instead.", FutureWarning
)
# pad expects a list of np.ndarray, but the previous feature extractors expected torch tensors
images = [to_numpy_array(image) for image in pixel_values_list]
return self.pad(
images=images,
return_pixel_mask=True,
return_tensors=return_tensors,
data_format=data_format,
)
def preprocess(
self,
images: ImageInput,

57
src/transformers/models/conditional_detr/image_processing_conditional_detr.py
View File

@@ -820,15 +820,6 @@ class ConditionalDetrImageProcessor(BaseImageProcessor):
self.image_std = image_std if image_std is not None else IMAGENET_DEFAULT_STD
self.do_pad = do_pad
@property
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.max_size
def max_size(self):
logger.warning(
"The `max_size` parameter is deprecated and will be removed in v4.27. "
"Please specify in `size['longest_edge'] instead`.",
)
return self.size["longest_edge"]
@classmethod
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.from_dict with Detr->ConditionalDetr
def from_dict(cls, image_processor_dict: Dict[str, Any], **kwargs):
@@ -873,7 +864,7 @@ class ConditionalDetrImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare
def prepare(self, image, target, return_segmentation_masks=False, masks_path=None):
logger.warning_once(
"The `prepare` method is deprecated and will be removed in a future version. "
"The `prepare` method is deprecated and will be removed in a v4.33. "
"Please use `prepare_annotation` instead. Note: the `prepare_annotation` method "
"does not return the image anymore.",
)
@@ -882,23 +873,17 @@ class ConditionalDetrImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.convert_coco_poly_to_mask
def convert_coco_poly_to_mask(self, *args, **kwargs):
logger.warning_once(
"The `convert_coco_poly_to_mask` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `convert_coco_poly_to_mask` method is deprecated and will be removed in v4.33. ")
return convert_coco_poly_to_mask(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_detection with DETR->ConditionalDetr
def prepare_coco_detection(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_detection` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_detection` method is deprecated and will be removed in v4.33. ")
return prepare_coco_detection_annotation(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_panoptic
def prepare_coco_panoptic(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_panoptic` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_panoptic` method is deprecated and will be removed in v4.33. ")
return prepare_coco_panoptic_annotation(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.resize
@@ -979,40 +964,6 @@ class ConditionalDetrImageProcessor(BaseImageProcessor):
"""
return normalize_annotation(annotation, image_size=image_size)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.pad_and_create_pixel_mask
def pad_and_create_pixel_mask(
self,
pixel_values_list: List[ImageInput],
return_tensors: Optional[Union[str, TensorType]] = None,
data_format: Optional[ChannelDimension] = None,
) -> BatchFeature:
"""
Pads a batch of images with zeros to the size of largest height and width in the batch and returns their
corresponding pixel mask.
Args:
images (`List[np.ndarray]`):
Batch of images to pad.
return_tensors (`str` or `TensorType`, *optional*):
The type of tensors to return. Can be one of:
- Unset: Return a list of `np.ndarray`.
- `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
- `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
- `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
- `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
data_format (`str` or `ChannelDimension`, *optional*):
The channel dimension format of the image. If not provided, it will be the same as the input image.
"""
logger.warning_once("This method is deprecated and will be removed in v4.27.0. Please use pad instead.")
# pad expects a list of np.ndarray, but the previous feature extractors expected torch tensors
images = [to_numpy_array(image) for image in pixel_values_list]
return self.pad(
images=images,
return_pixel_mask=True,
return_tensors=return_tensors,
data_format=data_format,
)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor._pad_image
def _pad_image(
self,

57
src/transformers/models/deformable_detr/image_processing_deformable_detr.py
View File

@@ -818,15 +818,6 @@ class DeformableDetrImageProcessor(BaseImageProcessor):
self.image_std = image_std if image_std is not None else IMAGENET_DEFAULT_STD
self.do_pad = do_pad
@property
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.max_size
def max_size(self):
logger.warning(
"The `max_size` parameter is deprecated and will be removed in v4.27. "
"Please specify in `size['longest_edge'] instead`.",
)
return self.size["longest_edge"]
@classmethod
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.from_dict with Detr->DeformableDetr
def from_dict(cls, image_processor_dict: Dict[str, Any], **kwargs):
@@ -871,7 +862,7 @@ class DeformableDetrImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare
def prepare(self, image, target, return_segmentation_masks=None, masks_path=None):
logger.warning_once(
"The `prepare` method is deprecated and will be removed in a future version. "
"The `prepare` method is deprecated and will be removed in a v4.33. "
"Please use `prepare_annotation` instead. Note: the `prepare_annotation` method "
"does not return the image anymore.",
)
@@ -880,23 +871,17 @@ class DeformableDetrImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.convert_coco_poly_to_mask
def convert_coco_poly_to_mask(self, *args, **kwargs):
logger.warning_once(
"The `convert_coco_poly_to_mask` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `convert_coco_poly_to_mask` method is deprecated and will be removed in v4.33. ")
return convert_coco_poly_to_mask(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_detection
def prepare_coco_detection(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_detection` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_detection` method is deprecated and will be removed in v4.33. ")
return prepare_coco_detection_annotation(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_panoptic
def prepare_coco_panoptic(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_panoptic` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_panoptic` method is deprecated and will be removed in v4.33. ")
return prepare_coco_panoptic_annotation(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.resize
@@ -977,40 +962,6 @@ class DeformableDetrImageProcessor(BaseImageProcessor):
"""
return normalize_annotation(annotation, image_size=image_size)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.pad_and_create_pixel_mask
def pad_and_create_pixel_mask(
self,
pixel_values_list: List[ImageInput],
return_tensors: Optional[Union[str, TensorType]] = None,
data_format: Optional[ChannelDimension] = None,
) -> BatchFeature:
"""
Pads a batch of images with zeros to the size of largest height and width in the batch and returns their
corresponding pixel mask.
Args:
images (`List[np.ndarray]`):
Batch of images to pad.
return_tensors (`str` or `TensorType`, *optional*):
The type of tensors to return. Can be one of:
- Unset: Return a list of `np.ndarray`.
- `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
- `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
- `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
- `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
data_format (`str` or `ChannelDimension`, *optional*):
The channel dimension format of the image. If not provided, it will be the same as the input image.
"""
logger.warning_once("This method is deprecated and will be removed in v4.27.0. Please use pad instead.")
# pad expects a list of np.ndarray, but the previous feature extractors expected torch tensors
images = [to_numpy_array(image) for image in pixel_values_list]
return self.pad(
images=images,
return_pixel_mask=True,
return_tensors=return_tensors,
data_format=data_format,
)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor._pad_image
def _pad_image(
self,

48
src/transformers/models/deta/image_processing_deta.py
View File

@@ -544,7 +544,7 @@ class DetaImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare
def prepare(self, image, target, return_segmentation_masks=None, masks_path=None):
logger.warning_once(
"The `prepare` method is deprecated and will be removed in a future version. "
"The `prepare` method is deprecated and will be removed in a v4.33. "
"Please use `prepare_annotation` instead. Note: the `prepare_annotation` method "
"does not return the image anymore.",
)
@@ -553,23 +553,17 @@ class DetaImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.convert_coco_poly_to_mask
def convert_coco_poly_to_mask(self, *args, **kwargs):
logger.warning_once(
"The `convert_coco_poly_to_mask` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `convert_coco_poly_to_mask` method is deprecated and will be removed in v4.33. ")
return convert_coco_poly_to_mask(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_detection
def prepare_coco_detection(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_detection` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_detection` method is deprecated and will be removed in v4.33. ")
return prepare_coco_detection_annotation(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_panoptic
def prepare_coco_panoptic(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_panoptic` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_panoptic` method is deprecated and will be removed in v4.33. ")
return prepare_coco_panoptic_annotation(*args, **kwargs)
def resize(
@@ -641,40 +635,6 @@ class DetaImageProcessor(BaseImageProcessor):
"""
return normalize_annotation(annotation, image_size=image_size)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.pad_and_create_pixel_mask
def pad_and_create_pixel_mask(
self,
pixel_values_list: List[ImageInput],
return_tensors: Optional[Union[str, TensorType]] = None,
data_format: Optional[ChannelDimension] = None,
) -> BatchFeature:
"""
Pads a batch of images with zeros to the size of largest height and width in the batch and returns their
corresponding pixel mask.
Args:
images (`List[np.ndarray]`):
Batch of images to pad.
return_tensors (`str` or `TensorType`, *optional*):
The type of tensors to return. Can be one of:
- Unset: Return a list of `np.ndarray`.
- `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
- `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
- `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
- `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
data_format (`str` or `ChannelDimension`, *optional*):
The channel dimension format of the image. If not provided, it will be the same as the input image.
"""
logger.warning_once("This method is deprecated and will be removed in v4.27.0. Please use pad instead.")
# pad expects a list of np.ndarray, but the previous feature extractors expected torch tensors
images = [to_numpy_array(image) for image in pixel_values_list]
return self.pad(
images=images,
return_pixel_mask=True,
return_tensors=return_tensors,
data_format=data_format,
)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor._pad_image
def _pad_image(
self,

55
src/transformers/models/detr/image_processing_detr.py
View File

@@ -802,14 +802,6 @@ class DetrImageProcessor(BaseImageProcessor):
self.image_std = image_std if image_std is not None else IMAGENET_DEFAULT_STD
self.do_pad = do_pad
@property
def max_size(self):
logger.warning(
"The `max_size` parameter is deprecated and will be removed in v4.27. "
"Please specify in `size['longest_edge'] instead`.",
)
return self.size["longest_edge"]
@classmethod
def from_dict(cls, image_processor_dict: Dict[str, Any], **kwargs):
"""
@@ -851,7 +843,7 @@ class DetrImageProcessor(BaseImageProcessor):
def prepare(self, image, target, return_segmentation_masks=None, masks_path=None):
logger.warning_once(
"The `prepare` method is deprecated and will be removed in a future version. "
"The `prepare` method is deprecated and will be removed in a v4.33. "
"Please use `prepare_annotation` instead. Note: the `prepare_annotation` method "
"does not return the image anymore.",
)
@@ -859,21 +851,15 @@ class DetrImageProcessor(BaseImageProcessor):
return image, target
def convert_coco_poly_to_mask(self, *args, **kwargs):
logger.warning_once(
"The `convert_coco_poly_to_mask` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `convert_coco_poly_to_mask` method is deprecated and will be removed in v4.33. ")
return convert_coco_poly_to_mask(*args, **kwargs)
def prepare_coco_detection(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_detection` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_detection` method is deprecated and will be removed in v4.33. ")
return prepare_coco_detection_annotation(*args, **kwargs)
def prepare_coco_panoptic(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_panoptic` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_panoptic` method is deprecated and will be removed in v4.33. ")
return prepare_coco_panoptic_annotation(*args, **kwargs)
def resize(
@@ -949,39 +935,6 @@ class DetrImageProcessor(BaseImageProcessor):
"""
return normalize_annotation(annotation, image_size=image_size)
def pad_and_create_pixel_mask(
self,
pixel_values_list: List[ImageInput],
return_tensors: Optional[Union[str, TensorType]] = None,
data_format: Optional[ChannelDimension] = None,
) -> BatchFeature:
"""
Pads a batch of images with zeros to the size of largest height and width in the batch and returns their
corresponding pixel mask.
Args:
images (`List[np.ndarray]`):
Batch of images to pad.
return_tensors (`str` or `TensorType`, *optional*):
The type of tensors to return. Can be one of:
- Unset: Return a list of `np.ndarray`.
- `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
- `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
- `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
- `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
data_format (`str` or `ChannelDimension`, *optional*):
The channel dimension format of the image. If not provided, it will be the same as the input image.
"""
logger.warning_once("This method is deprecated and will be removed in v4.27.0. Please use pad instead.")
# pad expects a list of np.ndarray, but the previous feature extractors expected torch tensors
images = [to_numpy_array(image) for image in pixel_values_list]
return self.pad(
images=images,
return_pixel_mask=True,
return_tensors=return_tensors,
data_format=data_format,
)
def _pad_image(
self,
image: np.ndarray,

6
src/transformers/models/donut/image_processing_donut.py
View File

@@ -151,12 +151,6 @@ class DonutImageProcessor(BaseImageProcessor):
return image
def rotate_image(self, *args, **kwargs):
logger.info(
"rotate_image is deprecated and will be removed in version 4.27. Please use align_long_axis instead."
)
return self.align_long_axis(*args, **kwargs)
def pad_image(
self,
image: np.ndarray,

27
src/transformers/models/mask2former/image_processing_mask2former.py
View File

@@ -29,7 +29,6 @@ from ...image_transforms import (
rescale,
resize,
to_channel_dimension_format,
to_numpy_array,
)
from ...image_utils import (
ChannelDimension,
@@ -38,6 +37,7 @@ from ...image_utils import (
get_image_size,
infer_channel_dimension_format,
is_batched,
to_numpy_array,
valid_images,
)
from ...utils import (
@@ -441,24 +441,6 @@ class Mask2FormerImageProcessor(BaseImageProcessor):
image_processor_dict["size_divisibility"] = kwargs.pop("size_divisibility")
return super().from_dict(image_processor_dict, **kwargs)
@property
def size_divisibility(self):
warnings.warn(
"The `size_divisibility` property is deprecated and will be removed in v4.27. Please use "
"`size_divisor` instead.",
FutureWarning,
)
return self.size_divisor
@property
def max_size(self):
warnings.warn(
"The `max_size` property is deprecated and will be removed in v4.27. Please use size['longest_edge']"
" instead.",
FutureWarning,
)
return self.size["longest_edge"]
def resize(
self,
image: np.ndarray,
@@ -789,7 +771,6 @@ class Mask2FormerImageProcessor(BaseImageProcessor):
ignore_index: Optional[int] = None,
reduce_labels: bool = False,
return_tensors: Optional[Union[str, TensorType]] = None,
**kwargs,
):
"""
Pad images up to the largest image in a batch and create a corresponding `pixel_mask`.
@@ -840,12 +821,6 @@ class Mask2FormerImageProcessor(BaseImageProcessor):
"""
ignore_index = self.ignore_index if ignore_index is None else ignore_index
reduce_labels = self.reduce_labels if reduce_labels is None else reduce_labels
if "pad_and_return_pixel_mask" in kwargs:
warnings.warn(
"The `pad_and_return_pixel_mask` argument has no effect and will be removed in v4.27", FutureWarning
)
pixel_values_list = [to_numpy_array(pixel_values) for pixel_values in pixel_values_list]
encoded_inputs = self.pad(pixel_values_list, return_tensors=return_tensors)

34
src/transformers/models/maskformer/image_processing_maskformer.py
View File

@@ -29,7 +29,6 @@ from ...image_transforms import (
rescale,
resize,
to_channel_dimension_format,
to_numpy_array,
)
from ...image_utils import (
ChannelDimension,
@@ -38,6 +37,7 @@ from ...image_utils import (
get_image_size,
infer_channel_dimension_format,
make_list_of_images,
to_numpy_array,
valid_images,
)
from ...utils import (
@@ -452,33 +452,6 @@ class MaskFormerImageProcessor(BaseImageProcessor):
image_processor_dict["size_divisibility"] = kwargs.pop("size_divisibility")
return super().from_dict(image_processor_dict, **kwargs)
@property
def size_divisibility(self):
warnings.warn(
"The `size_divisibility` property is deprecated and will be removed in v4.27. Please use "
"`size_divisor` instead.",
FutureWarning,
)
return self.size_divisor
@property
def max_size(self):
warnings.warn(
"The `max_size` property is deprecated and will be removed in v4.27. Please use size['longest_edge']"
" instead.",
FutureWarning,
)
return self.size["longest_edge"]
@property
def reduce_labels(self):
warnings.warn(
"The `reduce_labels` property is deprecated and will be removed in v4.27. Please use "
"`do_reduce_labels` instead.",
FutureWarning,
)
return self.do_reduce_labels
def resize(
self,
image: np.ndarray,
@@ -820,7 +793,6 @@ class MaskFormerImageProcessor(BaseImageProcessor):
ignore_index: Optional[int] = None,
reduce_labels: bool = False,
return_tensors: Optional[Union[str, TensorType]] = None,
**kwargs,
):
"""
Pad images up to the largest image in a batch and create a corresponding `pixel_mask`.
@@ -869,10 +841,6 @@ class MaskFormerImageProcessor(BaseImageProcessor):
`annotations` are provided). They identify the labels of `mask_labels`, e.g. the label of
`mask_labels[i][j]` if `class_labels[i][j]`.
"""
if "pad_and_return_pixel_mask" in kwargs:
warnings.warn(
"The `pad_and_return_pixel_mask` argument has no effect and will be removed in v4.27", FutureWarning
)
ignore_index = self.ignore_index if ignore_index is None else ignore_index
reduce_labels = self.do_reduce_labels if reduce_labels is None else reduce_labels

8
src/transformers/models/oneformer/image_processing_oneformer.py
View File

@@ -30,7 +30,6 @@ from ...image_transforms import (
rescale,
resize,
to_channel_dimension_format,
to_numpy_array,
)
from ...image_utils import (
ChannelDimension,
@@ -39,6 +38,7 @@ from ...image_utils import (
get_image_size,
infer_channel_dimension_format,
make_list_of_images,
to_numpy_array,
valid_images,
)
from ...utils import (
@@ -881,7 +881,6 @@ class OneFormerImageProcessor(BaseImageProcessor):
ignore_index: Optional[int] = None,
reduce_labels: bool = False,
return_tensors: Optional[Union[str, TensorType]] = None,
**kwargs,
):
"""
Pad images up to the largest image in a batch and create a corresponding `pixel_mask`.
@@ -935,11 +934,6 @@ class OneFormerImageProcessor(BaseImageProcessor):
- **text_inputs** -- Optional list of text string entries to be fed to a model (when `annotations` are
provided). They identify the binary masks present in the image.
"""
if "pad_and_return_pixel_mask" in kwargs:
warnings.warn(
"The `pad_and_return_pixel_mask` argument has no effect and will be removed in v4.27", FutureWarning
)
ignore_index = self.ignore_index if ignore_index is None else ignore_index
reduce_labels = self.do_reduce_labels if reduce_labels is None else reduce_labels
pixel_values_list = [to_numpy_array(pixel_values) for pixel_values in pixel_values_list]

2
src/transformers/models/owlvit/image_processing_owlvit.py
View File

@@ -27,7 +27,6 @@ from ...image_transforms import (
rescale,
resize,
to_channel_dimension_format,
to_numpy_array,
)
from ...image_utils import (
OPENAI_CLIP_MEAN,
@@ -36,6 +35,7 @@ from ...image_utils import (
ImageInput,
PILImageResampling,
make_list_of_images,
to_numpy_array,
valid_images,
)
from ...utils import TensorType, is_torch_available, logging

9
src/transformers/models/segformer/image_processing_segformer.py
View File

@@ -116,15 +116,6 @@ class SegformerImageProcessor(BaseImageProcessor):
self.image_std = image_std if image_std is not None else IMAGENET_DEFAULT_STD
self.do_reduce_labels = do_reduce_labels
@property
def reduce_labels(self):
warnings.warn(
"The `reduce_labels` property is deprecated and will be removed in a v4.27. Please use "
"`do_reduce_labels` instead.",
FutureWarning,
)
return self.do_reduce_labels
@classmethod
def from_dict(cls, image_processor_dict: Dict[str, Any], **kwargs):
"""

36
src/transformers/models/vilt/image_processing_vilt.py
View File

@@ -14,7 +14,6 @@
# limitations under the License.
"""Image processor class for Vilt."""
import warnings
from typing import Any, Dict, Iterable, List, Optional, Tuple, Union
import numpy as np
@@ -331,41 +330,6 @@ class ViltImageProcessor(BaseImageProcessor):
return BatchFeature(data=data, tensor_type=return_tensors)
def pad_and_create_pixel_mask(
self,
pixel_values_list: List[ImageInput],
return_tensors: Optional[Union[str, TensorType]] = None,
data_format: Optional[ChannelDimension] = None,
) -> BatchFeature:
"""
Pads a batch of images with zeros to the size of largest height and width in the batch and returns their
corresponding pixel mask.
Args:
images (`List[np.ndarray]`):
Batch of images to pad.
return_tensors (`str` or `TensorType`, *optional*):
The type of tensors to return. Can be one of:
- Unset: Return a list of `np.ndarray`.
- `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
- `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
- `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
- `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
data_format (`str` or `ChannelDimension`, *optional*):
The channel dimension format of the image. If not provided, it will be the same as the input image.
"""
warnings.warn(
"This method is deprecated and will be removed in v4.26.0. Please use pad instead.", FutureWarning
)
# pad expects a list of np.ndarray, but the previous feature extractors expected torch tensors
images = [to_numpy_array(image) for image in pixel_values_list]
return self.pad(
images=images,
return_pixel_mask=True,
return_tensors=return_tensors,
data_format=data_format,
)
def preprocess(
self,
images: ImageInput,

27
src/transformers/models/yolos/image_processing_yolos.py
View File

@@ -729,15 +729,6 @@ class YolosImageProcessor(BaseImageProcessor):
self.image_std = image_std if image_std is not None else IMAGENET_DEFAULT_STD
self.do_pad = do_pad
@property
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.max_size
def max_size(self):
logger.warning(
"The `max_size` parameter is deprecated and will be removed in v4.27. "
"Please specify in `size['longest_edge'] instead`.",
)
return self.size["longest_edge"]
@classmethod
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.from_dict with Detr->Yolos
def from_dict(cls, image_processor_dict: Dict[str, Any], **kwargs):
@@ -782,7 +773,7 @@ class YolosImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare
def prepare(self, image, target, return_segmentation_masks=False, masks_path=None):
logger.warning_once(
"The `prepare` method is deprecated and will be removed in a future version. "
"The `prepare` method is deprecated and will be removed in a v4.33. "
"Please use `prepare_annotation` instead. Note: the `prepare_annotation` method "
"does not return the image anymore.",
)
@@ -791,23 +782,17 @@ class YolosImageProcessor(BaseImageProcessor):
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.convert_coco_poly_to_mask
def convert_coco_poly_to_mask(self, *args, **kwargs):
logger.warning_once(
"The `convert_coco_poly_to_mask` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `convert_coco_poly_to_mask` method is deprecated and will be removed in v4.33. ")
return convert_coco_poly_to_mask(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_detection with DETR->Yolos
def prepare_coco_detection(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_detection` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_detection` method is deprecated and will be removed in v4.33. ")
return prepare_coco_detection_annotation(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.prepare_coco_panoptic
def prepare_coco_panoptic(self, *args, **kwargs):
logger.warning_once(
"The `prepare_coco_panoptic` method is deprecated and will be removed in a future version. "
)
logger.warning_once("The `prepare_coco_panoptic` method is deprecated and will be removed in v4.33. ")
return prepare_coco_panoptic_annotation(*args, **kwargs)
# Copied from transformers.models.detr.image_processing_detr.DetrImageProcessor.resize
@@ -1010,7 +995,7 @@ class YolosImageProcessor(BaseImageProcessor):
"""
if "pad_and_return_pixel_mask" in kwargs:
logger.warning_once(
"The `pad_and_return_pixel_mask` argument is deprecated and will be removed in a future version, "
"The `pad_and_return_pixel_mask` argument is deprecated and will be removed in v4.33, "
"use `do_pad` instead.",
)
do_pad = kwargs.pop("pad_and_return_pixel_mask")
@@ -1018,7 +1003,7 @@ class YolosImageProcessor(BaseImageProcessor):
max_size = None
if "max_size" in kwargs:
logger.warning_once(
"The `max_size` argument is deprecated and will be removed in a future version, use"
"The `max_size` argument is deprecated and will be removed in v4.33, use"
" `size['longest_edge']` instead.",
)
size = kwargs.pop("max_size")

20
tests/models/bridgetower/test_image_processing_bridgetower.py
View File

@@ -236,23 +236,3 @@ class BridgeTowerImageProcessingTest(ImageProcessingSavingTestMixin, unittest.Te
expected_width,
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image processors
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(do_resize=False, do_normalize=False, do_rescale=False)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.pad_and_create_pixel_mask(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)

20
tests/models/conditional_detr/test_image_processing_conditional_detr.py
View File

@@ -244,26 +244,6 @@ class ConditionalDetrImageProcessingTest(ImageProcessingSavingTestMixin, unittes
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processings
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(do_resize=False, do_normalize=False, do_rescale=False)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.pad_and_create_pixel_mask(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
@slow
def test_call_pytorch_with_coco_detection_annotations(self):
# prepare image and target

21
tests/models/deformable_detr/test_image_processing_deformable_detr.py
View File

@@ -246,27 +246,6 @@ class DeformableDetrImageProcessingTest(ImageProcessingSavingTestMixin, unittest
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processings
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(do_resize=False, do_normalize=False, do_rescale=False)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.pad_and_create_pixel_mask(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
@slow
def test_call_pytorch_with_coco_detection_annotations(self):
# prepare image and target

21
tests/models/deta/test_image_processing_deta.py
View File

@@ -240,27 +240,6 @@ class DetaImageProcessingTest(ImageProcessingSavingTestMixin, unittest.TestCase)
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processings
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(do_resize=False, do_normalize=False, do_rescale=False)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.pad_and_create_pixel_mask(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
@slow
def test_call_pytorch_with_coco_detection_annotations(self):
# prepare image and target

20
tests/models/detr/test_image_processing_detr.py
View File

@@ -247,26 +247,6 @@ class DetrImageProcessingTest(ImageProcessingSavingTestMixin, unittest.TestCase)
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processings
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(do_resize=False, do_normalize=False, do_rescale=False)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.pad_and_create_pixel_mask(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
@slow
def test_call_pytorch_with_coco_detection_annotations(self):
# prepare image and target

23
tests/models/mask2former/test_image_processing_mask2former.py
View File

@@ -147,7 +147,6 @@ class Mask2FormerImageProcessingTest(ImageProcessingSavingTestMixin, unittest.Te
self.assertTrue(hasattr(image_processing, "do_normalize"))
self.assertTrue(hasattr(image_processing, "do_resize"))
self.assertTrue(hasattr(image_processing, "size"))
self.assertTrue(hasattr(image_processing, "max_size"))
self.assertTrue(hasattr(image_processing, "ignore_index"))
self.assertTrue(hasattr(image_processing, "num_labels"))
@@ -263,28 +262,6 @@ class Mask2FormerImageProcessingTest(ImageProcessingSavingTestMixin, unittest.Te
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processings
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(
do_resize=False, do_normalize=False, do_rescale=False, num_labels=self.image_processor_tester.num_classes
)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.encode_inputs(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
def comm_get_image_processing_inputs(
self, with_segmentation_maps=False, is_instance_map=False, segmentation_type="np"
):

23
tests/models/maskformer/test_image_processing_maskformer.py
View File

@@ -147,7 +147,6 @@ class MaskFormerImageProcessingTest(ImageProcessingSavingTestMixin, unittest.Tes
self.assertTrue(hasattr(image_processing, "do_normalize"))
self.assertTrue(hasattr(image_processing, "do_resize"))
self.assertTrue(hasattr(image_processing, "size"))
self.assertTrue(hasattr(image_processing, "max_size"))
self.assertTrue(hasattr(image_processing, "ignore_index"))
self.assertTrue(hasattr(image_processing, "num_labels"))
@@ -263,28 +262,6 @@ class MaskFormerImageProcessingTest(ImageProcessingSavingTestMixin, unittest.Tes
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processings
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(
do_resize=False, do_normalize=False, do_rescale=False, num_labels=self.image_processor_tester.num_classes
)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.encode_inputs(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
def comm_get_image_processing_inputs(
self, with_segmentation_maps=False, is_instance_map=False, segmentation_type="np"
):

30
tests/models/oneformer/test_image_processing_oneformer.py
View File

@@ -286,36 +286,6 @@ class OneFormerImageProcessingTest(ImageProcessingSavingTestMixin, unittest.Test
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processors
image_processor_1 = self.image_processing_class(**self.image_processor_dict)
image_processor_2 = self.image_processing_class(
do_resize=False,
do_normalize=False,
do_rescale=False,
num_labels=self.image_processing_tester.num_classes,
class_info_file="ade20k_panoptic.json",
num_text=self.image_processing_tester.num_text,
repo_path="shi-labs/oneformer_demo",
)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processing_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processor_1.encode_inputs(
image_inputs, ["semantic"] * len(image_inputs), return_tensors="pt"
)
encoded_images = image_processor_2(image_inputs, ["semantic"] * len(image_inputs), return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
def comm_get_image_processor_inputs(
self, with_segmentation_maps=False, is_instance_map=False, segmentation_type="np"
):

35
tests/models/oneformer/test_processor_oneformer.py
View File

@@ -355,41 +355,6 @@ class OneFormerProcessingTest(unittest.TestCase):
(self.processing_tester.batch_size, expected_sequence_length),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize processors
processor_1 = self.processing_class(**self.processor_dict)
image_processor = OneFormerImageProcessor(
do_resize=False,
do_normalize=False,
do_rescale=False,
num_labels=self.processing_tester.num_classes,
class_info_file="ade20k_panoptic.json",
num_text=self.processing_tester.num_text,
)
tokenizer = CLIPTokenizer.from_pretrained("shi-labs/oneformer_ade20k_swin_tiny")
processor_2 = self.processing_class(
image_processor=image_processor, tokenizer=tokenizer, max_seq_length=77, task_seq_length=77
)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.processing_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = processor_1.encode_inputs(
image_inputs, ["semantic"] * len(image_inputs), return_tensors="pt"
)
encoded_images = processor_2(image_inputs, ["semantic"] * len(image_inputs), return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)
def comm_get_processor_inputs(self, with_segmentation_maps=False, is_instance_map=False, segmentation_type="np"):
processor = self.processing_class(**self.processor_dict)
# prepare image and target

20
tests/models/vilt/test_image_processing_vilt.py
View File

@@ -237,23 +237,3 @@ class ViltImageProcessingTest(ImageProcessingSavingTestMixin, unittest.TestCase)
expected_width,
),
)
def test_equivalence_pad_and_create_pixel_mask(self):
# Initialize image_processings
image_processing_1 = self.image_processing_class(**self.image_processor_dict)
image_processing_2 = self.image_processing_class(do_resize=False, do_normalize=False, do_rescale=False)
# create random PyTorch tensors
image_inputs = prepare_image_inputs(self.image_processor_tester, equal_resolution=False, torchify=True)
for image in image_inputs:
self.assertIsInstance(image, torch.Tensor)
# Test whether the method "pad_and_return_pixel_mask" and calling the image processor return the same tensors
encoded_images_with_method = image_processing_1.pad_and_create_pixel_mask(image_inputs, return_tensors="pt")
encoded_images = image_processing_2(image_inputs, return_tensors="pt")
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_values"], encoded_images["pixel_values"], atol=1e-4)
)
self.assertTrue(
torch.allclose(encoded_images_with_method["pixel_mask"], encoded_images["pixel_mask"], atol=1e-4)
)