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

Commitb4b4fea

Browse files
authored
Add efficientdet models (#67)
* add effdet check* Add effdet models and catalogs* clean-up* register effdet models* Add generalized image type support for layout models* Add effdet tests* Update reqs
1 parent9b73ff1 commitb4b4fea

File tree

12 files changed

+402
-24
lines changed

12 files changed

+402
-24
lines changed

‎dev-requirements.txt

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,5 @@
11
pytest
2+
torch
23
numpy
34
opencv-python
45
pandas
@@ -11,4 +12,5 @@ google-cloud-vision==1
1112
pytesseract
1213
pycocotools
1314
git+https://github.com/facebookresearch/detectron2.git@v0.4#egg=detectron2
14-
paddlepaddle
15+
paddlepaddle
16+
effdet

‎setup.py

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -24,7 +24,6 @@
2424
"pandas",
2525
"pillow",
2626
"pyyaml>=5.1",
27-
"torch",
2827
"torchvision",
2928
"iopath",
3029
],
@@ -33,6 +32,10 @@
3332
'google-cloud-vision==1',
3433
'pytesseract'
3534
],
35+
"effdet": [
36+
"torch",
37+
"effdet"
38+
]
3639
},
3740
include_package_data=True
3841
)

‎src/layoutparser/__init__.py

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -6,6 +6,7 @@
66
_LazyModule,
77
is_detectron2_available,
88
is_paddle_available,
9+
is_effdet_available,
910
is_pytesseract_available,
1011
is_gcv_available,
1112
)
@@ -45,6 +46,9 @@
4546
ifis_paddle_available():
4647
_import_structure["models.paddledetection"]= ["PaddleDetectionLayoutModel"]
4748

49+
ifis_effdet_available():
50+
_import_structure["models.effdet"]= ["EfficientDetLayoutModel"]
51+
4852
ifis_pytesseract_available():
4953
_import_structure["ocr.tesseract_agent"]= [
5054
"TesseractAgent",

‎src/layoutparser/file_utils.py

Lines changed: 22 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -39,11 +39,18 @@
3939
# The name of the paddlepaddle library:
4040
# Install name: pip install paddlepaddle
4141
# Import name: import paddle
42-
_paddle_version=importlib_metadata.version("paddlepaddle")
42+
_paddle_version=importlib_metadata.version("paddlepaddle")
4343
logger.debug(f"Paddle version{_paddle_version} available.")
4444
exceptimportlib_metadata.PackageNotFoundError:
4545
_paddle_available=False
4646

47+
_effdet_available=importlib.util.find_spec("effdet")isnotNone
48+
try:
49+
_effdet_version=importlib_metadata.version("effdet")
50+
logger.debug(f"Effdet version{_effdet_version} available.")
51+
exceptimportlib_metadata.PackageNotFoundError:
52+
_effdet_version=False
53+
4754
###########################################
4855
############## OCR Tool Deps ##############
4956
###########################################
@@ -78,12 +85,16 @@ def is_torch_cuda_available():
7885
returnFalse
7986

8087

88+
defis_detectron2_available():
89+
return_detectron2_available
90+
91+
8192
defis_paddle_available():
8293
return_paddle_available
8394

8495

85-
defis_detectron2_available():
86-
return_detectron2_available
96+
defis_effdet_available():
97+
return_effdet_available
8798

8899

89100
defis_pytesseract_available():
@@ -111,6 +122,11 @@ def is_gcv_available():
111122
installation page: https://github.com/PaddlePaddle/Paddle and follow the ones that match your environment.
112123
"""
113124

125+
EFFDET_IMPORT_ERROR="""
126+
{0} requires the effdet library but it was not found in your environment. You can install it with pip:
127+
`pip install effdet`
128+
"""
129+
114130
PYTESSERACT_IMPORT_ERROR="""
115131
{0} requires the PyTesseract library but it was not found in your environment. You can install it with pip:
116132
`pip install pytesseract`
@@ -126,6 +142,7 @@ def is_gcv_available():
126142
("torch", (is_torch_available,PYTORCH_IMPORT_ERROR)),
127143
("detectron2", (is_detectron2_available,DETECTRON2_IMPORT_ERROR)),
128144
("paddle", (is_paddle_available,PADDLE_IMPORT_ERROR)),
145+
("effdet", (is_effdet_available, )),
129146
("pytesseract", (is_pytesseract_available,PYTESSERACT_IMPORT_ERROR)),
130147
("google-cloud-vision", (is_gcv_available,GCV_IMPORT_ERROR)),
131148
]
@@ -172,7 +189,7 @@ def __init__(
172189
self._import_structure=import_structure
173190

174191
# Following [PEP 366](https://www.python.org/dev/peps/pep-0366/)
175-
# The __package__ variable should be set
192+
# The __package__ variable should be set
176193
# https://docs.python.org/3/reference/import.html#__package__
177194
self.__package__=self.__name__
178195

@@ -198,4 +215,4 @@ def _get_module(self, module_name: str):
198215
returnimportlib.import_module("."+module_name,self.__name__)
199216

200217
def__reduce__(self):
201-
return (self.__class__, (self._name,self.__file__,self._import_structure))
218+
return (self.__class__, (self._name,self.__file__,self._import_structure))

‎src/layoutparser/models/__init__.py

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,2 +1,3 @@
11
from .detectron2.layoutmodelimportDetectron2LayoutModel
2-
from .paddledetection.layoutmodelimportPaddleDetectionLayoutModel
2+
from .paddledetection.layoutmodelimportPaddleDetectionLayoutModel
3+
from .effdet.layoutmodelimportEfficientDetLayoutModel

‎src/layoutparser/models/base_layoutmodel.py

Lines changed: 9 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,17 +1,23 @@
1+
fromtypingimportUnion
12
fromabcimportABC,abstractmethod
23

34
from ..file_utilsimportrequires_backends
45

56

67
classBaseLayoutModel(ABC):
7-
88
@property
99
@abstractmethod
1010
defDETECTOR_NAME(self):
1111
pass
12-
12+
13+
@abstractmethod
14+
defdetect(self,image):
15+
pass
16+
1317
@abstractmethod
14-
defdetect(self):
18+
defimage_loader(self,image:Union["ndarray","Image"]):
19+
"""It will process the input images appropriately to the target format.
20+
"""
1521
pass
1622

1723
# Add lazy loading mechanisms for layout models, refer to

‎src/layoutparser/models/detectron2/layoutmodel.py

Lines changed: 10 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,4 @@
1+
fromtypingimportUnion
12
fromPILimportImage
23
importnumpyasnp
34

@@ -41,7 +42,7 @@ class Detectron2LayoutModel(BaseLayoutModel):
4142
4243
Examples::
4344
>>> import layoutparser as lp
44-
>>> model = lp.models.Detectron2LayoutModel('lp://HJDataset/faster_rcnn_R_50_FPN_3x/config')
45+
>>> model = lp.Detectron2LayoutModel('lp://HJDataset/faster_rcnn_R_50_FPN_3x/config')
4546
>>> model.detect(image)
4647
4748
"""
@@ -108,7 +109,7 @@ def _reconstruct_path_with_detector_name(self, path: str) -> str:
108109
model_name_segments=model_name.split("/")
109110
if (
110111
len(model_name_segments)==3
111-
and"detectron2"notinmodel_name_segments
112+
andself.DETECTOR_NAMEnotinmodel_name_segments
112113
):
113114
return"lp://"+self.DETECTOR_NAME+"/"+path[len("lp://") :]
114115
returnpath
@@ -148,12 +149,16 @@ def detect(self, image):
148149
:obj:`~layoutparser.Layout`: The detected layout of the input image
149150
"""
150151

152+
image=self.image_loader(image)
153+
outputs=self.model(image)
154+
layout=self.gather_output(outputs)
155+
returnlayout
156+
157+
defimage_loader(self,image:Union["np.ndarray","Image.Image"]):
151158
# Convert PIL Image Input
152159
ifisinstance(image,Image.Image):
153160
ifimage.mode!="RGB":
154161
image=image.convert("RGB")
155162
image=np.array(image)
156163

157-
outputs=self.model(image)
158-
layout=self.gather_output(outputs)
159-
returnlayout
164+
returnimage
Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,2 @@
1+
from .importcatalogas_UNUSED
2+
from .layoutmodelimportEfficientDetLayoutModel
Lines changed: 53 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,53 @@
1+
fromiopath.common.file_ioimportPathHandler
2+
3+
from ..base_catalogimportPathManager
4+
5+
MODEL_CATALOG= {
6+
"PubLayNet": {
7+
"tf_efficientdet_d0":"https://www.dropbox.com/s/ukbw5s673633hsw/publaynet-tf_efficientdet_d0.pth.tar?dl=1",
8+
"tf_efficientdet_d1":"https://www.dropbox.com/s/gxy11xkkiwnpgog/publaynet-tf_efficientdet_d1.pth.tar?dl=1"
9+
},
10+
"MFD": {
11+
"tf_efficientdet_d0":"https://www.dropbox.com/s/dkr22iux7thlhel/mfd-tf_efficientdet_d0.pth.tar?dl=1",
12+
"tf_efficientdet_d1":"https://www.dropbox.com/s/icmbiaqr5s9bz1x/mfd-tf_efficientdet_d1.pth.tar?dl=1"
13+
}
14+
}
15+
16+
# In effdet training scripts, it requires the label_map starting
17+
# from 1 instead of 0
18+
LABEL_MAP_CATALOG= {
19+
"PubLayNet": {
20+
1:"Text",
21+
2:"Title",
22+
3:"List",
23+
4:"Table",
24+
5:"Figure"
25+
}
26+
}
27+
28+
classLayoutParserEfficientDetModelHandler(PathHandler):
29+
"""
30+
Resolve anything that's in LayoutParser model zoo.
31+
"""
32+
33+
PREFIX="lp://efficientdet/"
34+
35+
def_get_supported_prefixes(self):
36+
return [self.PREFIX]
37+
38+
def_get_local_path(self,path,**kwargs):
39+
model_name=path[len(self.PREFIX) :]
40+
41+
dataset_name,*model_name,data_type=model_name.split("/")
42+
43+
ifdata_type=="weight":
44+
model_url=MODEL_CATALOG[dataset_name]["/".join(model_name)]
45+
else:
46+
raiseValueError(f"Unknown data_type{data_type}")
47+
returnPathManager.get_local_path(model_url,**kwargs)
48+
49+
def_open(self,path,mode="r",**kwargs):
50+
returnPathManager.open(self._get_local_path(path),mode,**kwargs)
51+
52+
53+
PathManager.register_handler(LayoutParserEfficientDetModelHandler())

0 commit comments

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