Skip to content

Type cast before normalize detr #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
amyeroberts merged 1 commit into from
Sep 2, 2022
Merged

Type cast before normalize detr #1

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 18 additions & 7 deletions src/transformers/models/detr/feature_extraction_detr.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -378,14 +378,14 @@ def get_size(image_size, size, max_size=None):

return rescaled_image, target

def _normalize(self, image, mean, std, target=None):
def _normalize(self, image, mean, std, target=None, rescale=False):
"""
Normalize the image with a certain mean and std.

If given, also normalize the target bounding boxes based on the size of the image.
"""

image = self.normalize(image, mean=mean, std=std)
image = self.normalize(image, mean=mean, std=std, rescale=rescale)
if target is None:
return image, None

Expand Down Expand Up @@ -455,9 +455,15 @@ def __call__(
- 1 for pixels that are real (i.e. **not masked**),
- 0 for pixels that are padding (i.e. **masked**).

return_tensors (`str` or [`~utils.TensorType`], *optional*):
If set, will return tensors instead of NumPy arrays. If set to `'pt'`, return PyTorch `torch.Tensor`
objects.
return_tensors (`str` or [`~utils.TensorType`], *optional*, defaults to `None`):
If set, will return a tensor of a particular framework.

Acceptable values are:
- `'tf'`: Return TensorFlow `tf.constant` objects.
- `'pt'`: Return PyTorch `torch.Tensor` objects.
- `'np'`: Return NumPy `np.ndarray` objects.
- `'jax'`: Return JAX `jnp.ndarray` objects.
- None: Return list of `np.ndarray` objects.

Returns:
[`BatchFeature`]: A [`BatchFeature`] with the following fields:
Expand Down Expand Up @@ -567,17 +573,22 @@ def __call__(
for idx, image in enumerate(images):
images[idx] = self._resize(image=image, target=None, size=self.size, max_size=self.max_size)[0]

# if do_normalize=False, the casting to a numpy array won't happen, so we need to do it here
make_channel_first = True if isinstance(images[0], Image.Image) else images[0].shape[-1] in (1, 3)
images = [self.to_numpy_array(image, rescale=False, channel_first=make_channel_first) for image in images]

if self.do_normalize:
if annotations is not None:
for idx, (image, target) in enumerate(zip(images, annotations)):
image, target = self._normalize(
image=image, mean=self.image_mean, std=self.image_std, target=target
image=image, mean=self.image_mean, std=self.image_std, target=target, rescale=True
)
images[idx] = image
annotations[idx] = target
else:
images = [
self._normalize(image=image, mean=self.image_mean, std=self.image_std)[0] for image in images
self._normalize(image=image, mean=self.image_mean, std=self.image_std, rescale=True)[0]
for image in images
]

if pad_and_return_pixel_mask:
Expand Down
41 changes: 41 additions & 0 deletions tests/models/detr/test_feature_extraction_detr.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,7 @@

import numpy as np

from parameterized import parameterized
from transformers.testing_utils import require_torch, require_vision, slow
from transformers.utils import is_torch_available, is_vision_available

Expand Down Expand Up @@ -336,3 +337,43 @@ def test_call_pytorch_with_coco_panoptic_annotations(self):
# verify size
expected_size = torch.tensor([800, 1066])
assert torch.allclose(encoding["labels"][0]["size"], expected_size)

@parameterized.expand(
[
("do_resize_True_do_normalize_True_pad_True", True, True, True),
("do_resize_True_do_normalize_False_pad_True", True, False, True),
("do_resize_False_do_normalize_False_pad_True", False, False, True),
("do_resize_False_do_normalize_True_pad_True", False, True, True),
("do_resize_True_do_normalize_True_pad_False", True, True, False),
("do_resize_True_do_normalize_False_pad_False", True, False, False),
("do_resize_False_do_normalize_False_pad_False", False, False, False),
("do_resize_False_do_normalize_True_pad_False", False, True, False),
]
)
def test_call_flags(self, _, do_resize, do_normalize, pad):
# Initialize feature_extractor
feature_extractor = self.feature_extraction_class(**self.feat_extract_dict)
feature_extractor.do_resize = do_resize
feature_extractor.do_normalize = do_normalize
# create random PIL images
image_inputs = prepare_image_inputs(self.feature_extract_tester, equal_resolution=False)

expected_shapes = [(3, *x.size[::-1]) for x in image_inputs]
if do_resize:
expected_shapes = [
(
self.feature_extract_tester.num_channels,
*self.feature_extract_tester.get_expected_values([image], batched=False),
)
for image in image_inputs
]
if pad:
expected_shapes = [tuple(max(x) for x in zip(*expected_shapes))] * len(image_inputs)

pixel_values = feature_extractor(image_inputs, pad_and_return_pixel_mask=pad, return_tensors=None)[
"pixel_values"
]
self.assertEqual(len(pixel_values), self.feature_extract_tester.batch_size)
for idx, image in enumerate(pixel_values):
self.assertEqual(image.shape, expected_shapes[idx])
self.assertIsInstance(image, np.ndarray)