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

Add Image Processors (#19796)

Browse Source 
* Add CLIP image processor

* Crop size as dict too

* Update warning

* Actually use logger this time

* Normalize doesn't change dtype of input

* Add perceiver image processor

* Tidy up

* Add DPT image processor

* Add Vilt image processor

* Tidy up

* Add poolformer image processor

* Tidy up

* Add LayoutLM v2 and v3 imsge processors

* Tidy up

* Add Flava image processor

* Tidy up

* Add deit image processor

* Tidy up

* Add ConvNext image processor

* Tidy up

* Add levit image processor

* Add segformer image processor

* Add in post processing

* Fix up

* Add ImageGPT image processor

* Fixup

* Add mobilevit image processor

* Tidy up

* Add postprocessing

* Fixup

* Add VideoMAE image processor

* Tidy up

* Add ImageGPT image processor

* Fixup

* Add ViT image processor

* Tidy up

* Add beit image processor

* Add mobilevit image processor

* Tidy up

* Add postprocessing

* Fixup

* Fix up

* Fix flava and remove tree module

* Fix image classification pipeline failing tests

* Update feature extractor in trainer scripts

* Update pad_if_smaller to accept tuple and int size

* Update for image segmentation pipeline

* Update src/transformers/models/perceiver/image_processing_perceiver.py

Co-authored-by: Alara Dirik <8944735+alaradirik@users.noreply.github.com>

* Update src/transformers/image_processing_utils.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/beit/image_processing_beit.py

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* PR comments - docstrings; remove accidentally added resize; var names

* Update docstrings

* Add exception if size is not in the right format

* Fix exception check

* Fix up

* Use shortest_edge in tuple in script

Co-authored-by: Alara Dirik <8944735+alaradirik@users.noreply.github.com>
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>
This commit is contained in:
amyeroberts
2022-11-02 11:57:36 +00:00
committed by GitHub
parent 2e3452af0f
commit a6b7759880
65 changed files with 7060 additions and 3590 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
examples/pytorch/image-classification/run_image_classification.py
View File

@@ -291,10 +291,14 @@ def main():
)
# Define torchvision transforms to be applied to each image.
if "shortest_edge" in feature_extractor.size:
size = feature_extractor.size["shortest_edge"]
else:
size = (feature_extractor.size["height"], feature_extractor.size["width"])
normalize = Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std)
_train_transforms = Compose(
[
RandomResizedCrop(feature_extractor.size),
RandomResizedCrop(size),
RandomHorizontalFlip(),
ToTensor(),
normalize,
@@ -302,8 +306,8 @@ def main():
)
_val_transforms = Compose(
[
Resize(feature_extractor.size),
CenterCrop(feature_extractor.size),
Resize(size),
CenterCrop(size),
ToTensor(),
normalize,
]

10
examples/pytorch/image-classification/run_image_classification_no_trainer.py
View File

@@ -315,10 +315,14 @@ def main():
# Preprocessing the datasets
# Define torchvision transforms to be applied to each image.
if "shortest_edge" in feature_extractor.size:
size = feature_extractor.size["shortest_edge"]
else:
size = (feature_extractor.size["height"], feature_extractor.size["width"])
normalize = Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std)
train_transforms = Compose(
[
RandomResizedCrop(feature_extractor.size),
RandomResizedCrop(size),
RandomHorizontalFlip(),
ToTensor(),
normalize,
@@ -326,8 +330,8 @@ def main():
)
val_transforms = Compose(
[
Resize(feature_extractor.size),
CenterCrop(feature_extractor.size),
Resize(size),
CenterCrop(size),
ToTensor(),
normalize,
]

6
examples/pytorch/image-pretraining/run_mae.py
View File

@@ -298,10 +298,14 @@ def main():
# transformations as done in original MAE paper
# source: https://github.com/facebookresearch/mae/blob/main/main_pretrain.py
if "shortest_edge" in feature_extractor.size:
size = feature_extractor.size["shortest_edge"]
else:
size = (feature_extractor.size["height"], feature_extractor.size["width"])
transforms = Compose(
[
Lambda(lambda img: img.convert("RGB") if img.mode != "RGB" else img),
RandomResizedCrop(feature_extractor.size, scale=(0.2, 1.0), interpolation=InterpolationMode.BICUBIC),
RandomResizedCrop(size, scale=(0.2, 1.0), interpolation=InterpolationMode.BICUBIC),
RandomHorizontalFlip(),
ToTensor(),
Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std),

26
examples/pytorch/semantic-segmentation/run_semantic_segmentation.py
View File

@@ -57,12 +57,11 @@ require_version("datasets>=2.0.0", "To fix: pip install -r examples/pytorch/sema
def pad_if_smaller(img, size, fill=0):
min_size = min(img.size)
if min_size < size:
original_width, original_height = img.size
pad_height = size - original_height if original_height < size else 0
pad_width = size - original_width if original_width < size else 0
img = functional.pad(img, (0, 0, pad_width, pad_height), fill=fill)
size = (size, size) if isinstance(size, int) else size
original_width, original_height = img.size
pad_height = size[1] - original_height if original_height < size[1] else 0
pad_width = size[0] - original_width if original_width < size[0] else 0
img = functional.pad(img, (0, 0, pad_width, pad_height), fill=fill)
return img
@@ -110,12 +109,12 @@ class RandomResize:
class RandomCrop:
def __init__(self, size):
self.size = size
self.size = size if isinstance(size, tuple) else (size, size)
def __call__(self, image, target):
image = pad_if_smaller(image, self.size)
target = pad_if_smaller(target, self.size, fill=255)
crop_params = transforms.RandomCrop.get_params(image, (self.size, self.size))
crop_params = transforms.RandomCrop.get_params(image, self.size)
image = functional.crop(image, *crop_params)
target = functional.crop(target, *crop_params)
return image, target
@@ -359,7 +358,7 @@ def main():
references=labels,
num_labels=len(id2label),
ignore_index=0,
reduce_labels=feature_extractor.reduce_labels,
reduce_labels=feature_extractor.do_reduce_labels,
)
# add per category metrics as individual key-value pairs
per_category_accuracy = metrics.pop("per_category_accuracy").tolist()
@@ -396,10 +395,15 @@ def main():
# Define torchvision transforms to be applied to each image + target.
# Not that straightforward in torchvision: https://github.com/pytorch/vision/issues/9
# Currently based on official torchvision references: https://github.com/pytorch/vision/blob/main/references/segmentation/transforms.py
if "shortest_edge" in feature_extractor.size:
# We instead set the target size as (shortest_edge, shortest_edge) to here to ensure all images are batchable.
size = (feature_extractor.size["shortest_edge"], feature_extractor.size["shortest_edge"])
else:
size = (feature_extractor.size["height"], feature_extractor.size["width"])
train_transforms = Compose(
[
ReduceLabels() if data_args.reduce_labels else Identity(),
RandomCrop(size=feature_extractor.size),
RandomCrop(size=size),
RandomHorizontalFlip(flip_prob=0.5),
PILToTensor(),
ConvertImageDtype(torch.float),
@@ -411,7 +415,7 @@ def main():
val_transforms = Compose(
[
ReduceLabels() if data_args.reduce_labels else Identity(),
Resize(size=(feature_extractor.size, feature_extractor.size)),
Resize(size=size),
PILToTensor(),
ConvertImageDtype(torch.float),
Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std),

9
examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py
View File

@@ -405,10 +405,15 @@ def main():
# Define torchvision transforms to be applied to each image + target.
# Not that straightforward in torchvision: https://github.com/pytorch/vision/issues/9
# Currently based on official torchvision references: https://github.com/pytorch/vision/blob/main/references/segmentation/transforms.py
if "shortest_edge" in feature_extractor.size:
# We instead set the target size as (shortest_edge, shortest_edge) to here to ensure all images are batchable.
size = (feature_extractor.size["shortest_edge"], feature_extractor.size["shortest_edge"])
else:
size = (feature_extractor.size["height"], feature_extractor.size["width"])
train_transforms = Compose(
[
ReduceLabels() if args.reduce_labels else Identity(),
RandomCrop(size=feature_extractor.size),
RandomCrop(size=size),
RandomHorizontalFlip(flip_prob=0.5),
PILToTensor(),
ConvertImageDtype(torch.float),
@@ -420,7 +425,7 @@ def main():
val_transforms = Compose(
[
ReduceLabels() if args.reduce_labels else Identity(),
Resize(size=(feature_extractor.size, feature_extractor.size)),
Resize(size=size),
PILToTensor(),
ConvertImageDtype(torch.float),
Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std),