Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commite70bf05

Browse files
authored
[fix] Improve dependencies for multi-backend support (#79)
* Add IMAGENET_DEFAULT_MEAN and IMAGENET_DEFAULT_STD* Try update backend-specific tests* modify test_effdet_backend ci* add test_only_effdet.py* more backend specific tests* Add pytest* escape torch.Tensor type annotation* misc fixes* install "torchvision" is more robust
1 parentb9fd396 commite70bf05

File tree

6 files changed

+161
-14
lines changed

6 files changed

+161
-14
lines changed

‎.github/workflows/ci.yml

Lines changed: 57 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -3,15 +3,68 @@ name: CI
33
on:[push]
44

55
jobs:
6-
build:
76

7+
test_only_effdet_backend:
8+
9+
runs-on:ubuntu-latest
10+
steps:
11+
-uses:actions/checkout@v2
12+
-uses:actions/setup-python@v2
13+
with:
14+
python-version:'3.7'
15+
16+
-name:Test Dependency Support
17+
run:|
18+
pip install pytest
19+
pip install -e . # The bare layoutparser module
20+
pytest tests_deps/test_file_utils.py
21+
22+
-name:Install only effdet deps
23+
run:|
24+
pip install pytest
25+
pip install -e ".[effdet]"
26+
pytest tests_deps/test_only_effdet.py
27+
28+
test_only_detectron2_backend:
29+
30+
runs-on:ubuntu-latest
31+
steps:
32+
-uses:actions/checkout@v2
33+
-uses:actions/setup-python@v2
34+
with:
35+
python-version:'3.7'
36+
37+
-name:Install only Detectron2 deps
38+
run:|
39+
pip install pytest
40+
pip install -e .
41+
pip install torchvision && pip install "git+https://github.com/facebookresearch/detectron2.git@v0.5#egg=detectron2"
42+
pytest tests_deps/test_only_detectron2.py
43+
44+
test_only_paddledetection_backend:
45+
46+
runs-on:ubuntu-latest
47+
steps:
48+
-uses:actions/checkout@v2
49+
-uses:actions/setup-python@v2
50+
with:
51+
python-version:'3.7'
52+
53+
-name:Install only PaddleDetection deps
54+
run:|
55+
pip install pytest
56+
pip install -e ".[paddledetection]"
57+
pytest tests_deps/test_only_paddledetection.py
58+
59+
test_all_methods_all_backends:
60+
needs:[test_only_effdet_backend, test_only_detectron2_backend, test_only_paddledetection_backend]
861
runs-on:ubuntu-latest
962
strategy:
1063
matrix:
1164
python-version:[3.7, 3.8]
12-
1365
steps:
1466
-uses:actions/checkout@v2
67+
1568
-name:Set up Python ${{ matrix.python-version }}
1669
uses:actions/setup-python@v2
1770
with:
@@ -29,16 +82,10 @@ jobs:
2982
flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics --ignore F821
3083
# exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
3184
flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
32-
33-
-name:Test Dependency Support
34-
run:|
35-
# Install additional requirements when running tests
36-
pip install pytest
37-
pytest tests_deps
3885
3986
-name:Test with pytest
4087
run:|
4188
# Install additional requirements when running tests
42-
pip install .[effdet]
89+
pip install".[effdet]"
4390
pip install -r dev-requirements.txt
44-
pytest tests
91+
pytest tests

‎installation.md

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -15,7 +15,7 @@ After several major updates, LayoutParser provides various functionalities and d
1515
| ---| ---|
1616
|`pip install layoutparser`|**Install the base LayoutParser Library**<br>It will support all key functions in LayoutParser, including:<br />1. Layout Data Structure and operations<br />2. Layout Visualization <br />3. Load/export the layout data|
1717
|`pip install "layoutparser[effdet]"`|**Install LayoutParser with Layout Detection Model Support**<br />It will install the LayoutParser base library as well as<br />supporting dependencies for the***EfficientDet***-based layout detection models.|
18-
|`pip install layoutparsertorch && pip install "git+https://github.com/facebookresearch/detectron2.git@v0.5#egg=detectron2"`|**Install LayoutParser with Layout Detection Model Support**<br />It will install the LayoutParser base library as well as<br />supporting dependencies for the***Detectron2***-based layout detection models. See details in[Additional Instruction: Install Detectron2 Layout Model Backend](#additional-instruction-install-detectron2-layout-model-backend).|
18+
|`pip install layoutparsertorchvision && pip install "git+https://github.com/facebookresearch/detectron2.git@v0.5#egg=detectron2"`|**Install LayoutParser with Layout Detection Model Support**<br />It will install the LayoutParser base library as well as<br />supporting dependencies for the***Detectron2***-based layout detection models. See details in[Additional Instruction: Install Detectron2 Layout Model Backend](#additional-instruction-install-detectron2-layout-model-backend).|
1919
|`pip install "layoutparser[paddledetection]"`|**Install LayoutParser with Layout Detection Model Support**<br />It will install the LayoutParser base library as well as<br />supporting dependencies for the***PaddleDetection***-based layout detection models.|
2020
|`pip install "layoutparser[ocr]"`|**Install LayoutParser with OCR Support**<br />It will install the LayoutParser base library as well as<br />supporting dependencies for performing OCRs. See details in[Additional Instruction: Install OCR utils](#additional-instruction-install-ocr-utils).|
2121

@@ -26,7 +26,7 @@ After several major updates, LayoutParser provides various functionalities and d
2626
If you would like to use the Detectron2 models for layout detection, you might need to run the following command:
2727

2828
```bash
29-
pip install layoutparsertorch&& pip install"detectron2@git+https://github.com/facebookresearch/detectron2.git@v0.5#egg=detectron2"
29+
pip install layoutparsertorchvision&& pip install"detectron2@git+https://github.com/facebookresearch/detectron2.git@v0.5#egg=detectron2"
3030
```
3131

3232
This might take some time as the command will*compile* the library. If you also want to install a Detectron2 version

‎src/layoutparser/models/effdet/layoutmodel.py

Lines changed: 9 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -31,6 +31,13 @@
3131
IMAGENET_DEFAULT_STD,
3232
transforms_coco_eval,
3333
)
34+
else:
35+
# Copied from https://github.com/rwightman/efficientdet-pytorch/blob/c5b694aa34900fdee6653210d856ca8320bf7d4e/effdet/data/transforms.py#L13
36+
# Such that when effdet is not loaded, we'll still have default values for IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
37+
IMAGENET_DEFAULT_MEAN= (0.485,0.456,0.406)
38+
IMAGENET_DEFAULT_STD= (0.229,0.224,0.225)
39+
# IMAGENET_INCEPTION_MEAN = (0.5, 0.5, 0.5)
40+
# IMAGENET_INCEPTION_STD = (0.5, 0.5, 0.5)
3441

3542

3643
classInputTransform:
@@ -56,7 +63,7 @@ def __init__(
5663
self.mean_tensor=torch.tensor([x*255forxinmean]).view(1,3,1,1)
5764
self.std_tensor=torch.tensor([x*255forxinstd]).view(1,3,1,1)
5865

59-
defpreprocess(self,image:Image)->Tuple[torch.Tensor,Dict]:
66+
defpreprocess(self,image:Image)->Tuple["torch.Tensor",Dict]:
6067

6168
image=image.convert("RGB")
6269
image_info= {"img_size":image.size}
@@ -206,7 +213,7 @@ def detect(self, image: Union["np.ndarray", "Image.Image"]):
206213
layout=self.gather_output(model_outputs)
207214
returnlayout
208215

209-
defgather_output(self,model_outputs:torch.Tensor)->Layout:
216+
defgather_output(self,model_outputs:"torch.Tensor")->Layout:
210217

211218
model_outputs=model_outputs.cpu().detach()
212219
box_predictions=Layout()

‎tests_deps/test_only_detectron2.py

Lines changed: 31 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,31 @@
1+
# Copyright 2021 The Layout Parser team. All rights reserved.
2+
#
3+
# Licensed under the Apache License, Version 2.0 (the "License");
4+
# you may not use this file except in compliance with the License.
5+
# You may obtain a copy of the License at
6+
#
7+
# http://www.apache.org/licenses/LICENSE-2.0
8+
#
9+
# Unless required by applicable law or agreed to in writing, software
10+
# distributed under the License is distributed on an "AS IS" BASIS,
11+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12+
# See the License for the specific language governing permissions and
13+
# limitations under the License.
14+
15+
importcv2
16+
importpytest
17+
fromlayoutparserimportDetectron2LayoutModel
18+
19+
deftest_only_effdet_model():
20+
21+
# When all the backeds are not installed, it should
22+
# elicit only ImportErrors
23+
24+
config="lp://PubLayNet/faster_rcnn_R_50_FPN_3x/config"
25+
model=Detectron2LayoutModel(config)
26+
image=cv2.imread("tests/fixtures/model/test_model_image.jpg")
27+
layout=model.detect(image)
28+
29+
withpytest.raises(ImportError):
30+
fromlayoutparserimportEfficientDetLayoutModel
31+
fromlayoutparserimportPaddleDetectionLayoutModel

‎tests_deps/test_only_effdet.py

Lines changed: 31 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,31 @@
1+
# Copyright 2021 The Layout Parser team. All rights reserved.
2+
#
3+
# Licensed under the Apache License, Version 2.0 (the "License");
4+
# you may not use this file except in compliance with the License.
5+
# You may obtain a copy of the License at
6+
#
7+
# http://www.apache.org/licenses/LICENSE-2.0
8+
#
9+
# Unless required by applicable law or agreed to in writing, software
10+
# distributed under the License is distributed on an "AS IS" BASIS,
11+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12+
# See the License for the specific language governing permissions and
13+
# limitations under the License.
14+
15+
importcv2
16+
importpytest
17+
fromlayoutparserimportEfficientDetLayoutModel
18+
19+
deftest_only_effdet_model():
20+
21+
# When all the backeds are not installed, it should
22+
# elicit only ImportErrors
23+
24+
config="lp://PubLayNet/tf_efficientdet_d0/config"
25+
model=EfficientDetLayoutModel(config)
26+
image=cv2.imread("tests/fixtures/model/test_model_image.jpg")
27+
layout=model.detect(image)
28+
29+
withpytest.raises(ImportError):
30+
fromlayoutparserimportDetectron2LayoutModel
31+
fromlayoutparserimportPaddleDetectionLayoutModel
Lines changed: 31 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,31 @@
1+
# Copyright 2021 The Layout Parser team. All rights reserved.
2+
#
3+
# Licensed under the Apache License, Version 2.0 (the "License");
4+
# you may not use this file except in compliance with the License.
5+
# You may obtain a copy of the License at
6+
#
7+
# http://www.apache.org/licenses/LICENSE-2.0
8+
#
9+
# Unless required by applicable law or agreed to in writing, software
10+
# distributed under the License is distributed on an "AS IS" BASIS,
11+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12+
# See the License for the specific language governing permissions and
13+
# limitations under the License.
14+
15+
importcv2
16+
importpytest
17+
fromlayoutparserimportPaddleDetectionLayoutModel
18+
19+
deftest_only_effdet_model():
20+
21+
# When all the backeds are not installed, it should
22+
# elicit only ImportErrors
23+
24+
config="lp://PubLayNet/ppyolov2_r50vd_dcn_365e/config"
25+
model=PaddleDetectionLayoutModel(config)
26+
image=cv2.imread("tests/fixtures/model/test_model_image.jpg")
27+
layout=model.detect(image)
28+
29+
withpytest.raises(ImportError):
30+
fromlayoutparserimportEfficientDetLayoutModel
31+
fromlayoutparserimportDetectron2LayoutModel

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp