提交 037e17fc 编写于 作者: W WenmuZhou

merge dygraph

上级 6127aad9
include LICENSE.txt
include LICENSE
include README.md
recursive-include ppocr/utils *.txt utility.py logging.py
recursive-include ppocr/utils *.txt utility.py logging.py network.py
recursive-include ppocr/data/ *.py
recursive-include ppocr/postprocess *.py
recursive-include tools/infer *.py
......
......@@ -19,17 +19,16 @@ __dir__ = os.path.dirname(__file__)
sys.path.append(os.path.join(__dir__, ''))
import cv2
import logging
import numpy as np
from pathlib import Path
import tarfile
import requests
from tqdm import tqdm
from tools.infer import predict_system
from ppocr.utils.logging import get_logger
logger = get_logger()
from ppocr.utils.utility import check_and_read_gif, get_image_file_list
from ppocr.utils.network import maybe_download, download_with_progressbar
from tools.infer.utility import draw_ocr, init_args, str2bool
__all__ = ['PaddleOCR']
......@@ -123,50 +122,6 @@ SUPPORT_REC_MODEL = ['CRNN']
BASE_DIR = os.path.expanduser("~/.paddleocr/")
def download_with_progressbar(url, save_path):
response = requests.get(url, stream=True)
total_size_in_bytes = int(response.headers.get('content-length', 0))
block_size = 1024 # 1 Kibibyte
progress_bar = tqdm(total=total_size_in_bytes, unit='iB', unit_scale=True)
with open(save_path, 'wb') as file:
for data in response.iter_content(block_size):
progress_bar.update(len(data))
file.write(data)
progress_bar.close()
if total_size_in_bytes == 0 or progress_bar.n != total_size_in_bytes:
logger.error("Something went wrong while downloading models")
sys.exit(0)
def maybe_download(model_storage_directory, url):
# using custom model
tar_file_name_list = [
'inference.pdiparams', 'inference.pdiparams.info', 'inference.pdmodel'
]
if not os.path.exists(
os.path.join(model_storage_directory, 'inference.pdiparams')
) or not os.path.exists(
os.path.join(model_storage_directory, 'inference.pdmodel')):
tmp_path = os.path.join(model_storage_directory, url.split('/')[-1])
print('download {} to {}'.format(url, tmp_path))
os.makedirs(model_storage_directory, exist_ok=True)
download_with_progressbar(url, tmp_path)
with tarfile.open(tmp_path, 'r') as tarObj:
for member in tarObj.getmembers():
filename = None
for tar_file_name in tar_file_name_list:
if tar_file_name in member.name:
filename = tar_file_name
if filename is None:
continue
file = tarObj.extractfile(member)
with open(
os.path.join(model_storage_directory, filename),
'wb') as f:
f.write(file.read())
os.remove(tmp_path)
def parse_args(mMain=True):
import argparse
parser = init_args()
......@@ -194,10 +149,12 @@ class PaddleOCR(predict_system.TextSystem):
args:
**kwargs: other params show in paddleocr --help
"""
postprocess_params = parse_args(mMain=False)
postprocess_params.__dict__.update(**kwargs)
self.use_angle_cls = postprocess_params.use_angle_cls
lang = postprocess_params.lang
params = parse_args(mMain=False)
params.__dict__.update(**kwargs)
if params.show_log:
logger.setLevel(logging.DEBUG)
self.use_angle_cls = params.use_angle_cls
lang = params.lang
latin_lang = [
'af', 'az', 'bs', 'cs', 'cy', 'da', 'de', 'es', 'et', 'fr', 'ga',
'hr', 'hu', 'id', 'is', 'it', 'ku', 'la', 'lt', 'lv', 'mi', 'ms',
......@@ -229,40 +186,40 @@ class PaddleOCR(predict_system.TextSystem):
else:
det_lang = "en"
use_inner_dict = False
if postprocess_params.rec_char_dict_path is None:
if params.rec_char_dict_path is None:
use_inner_dict = True
postprocess_params.rec_char_dict_path = model_urls['rec'][lang][
params.rec_char_dict_path = model_urls['rec'][lang][
'dict_path']
# init model dir
if postprocess_params.det_model_dir is None:
postprocess_params.det_model_dir = os.path.join(BASE_DIR, VERSION,
if params.det_model_dir is None:
params.det_model_dir = os.path.join(BASE_DIR, VERSION,
'det', det_lang)
if postprocess_params.rec_model_dir is None:
postprocess_params.rec_model_dir = os.path.join(BASE_DIR, VERSION,
if params.rec_model_dir is None:
params.rec_model_dir = os.path.join(BASE_DIR, VERSION,
'rec', lang)
if postprocess_params.cls_model_dir is None:
postprocess_params.cls_model_dir = os.path.join(BASE_DIR, 'cls')
print(postprocess_params)
if params.cls_model_dir is None:
params.cls_model_dir = os.path.join(BASE_DIR, 'cls')
# download model
maybe_download(postprocess_params.det_model_dir,
maybe_download(params.det_model_dir,
model_urls['det'][det_lang])
maybe_download(postprocess_params.rec_model_dir,
maybe_download(params.rec_model_dir,
model_urls['rec'][lang]['url'])
maybe_download(postprocess_params.cls_model_dir, model_urls['cls'])
maybe_download(params.cls_model_dir, model_urls['cls'])
if postprocess_params.det_algorithm not in SUPPORT_DET_MODEL:
if params.det_algorithm not in SUPPORT_DET_MODEL:
logger.error('det_algorithm must in {}'.format(SUPPORT_DET_MODEL))
sys.exit(0)
if postprocess_params.rec_algorithm not in SUPPORT_REC_MODEL:
if params.rec_algorithm not in SUPPORT_REC_MODEL:
logger.error('rec_algorithm must in {}'.format(SUPPORT_REC_MODEL))
sys.exit(0)
if use_inner_dict:
postprocess_params.rec_char_dict_path = str(
Path(__file__).parent / postprocess_params.rec_char_dict_path)
params.rec_char_dict_path = str(
Path(__file__).parent / params.rec_char_dict_path)
print(params)
# init det_model and rec_model
super().__init__(postprocess_params)
super().__init__(params)
def ocr(self, img, det=True, rec=True, cls=True):
"""
......
......@@ -163,7 +163,7 @@ class DetResizeForTest(object):
img, (ratio_h, ratio_w)
"""
limit_side_len = self.limit_side_len
h, w, _ = img.shape
h, w, c = img.shape
# limit the max side
if self.limit_type == 'max':
......@@ -174,7 +174,7 @@ class DetResizeForTest(object):
ratio = float(limit_side_len) / w
else:
ratio = 1.
else:
elif self.limit_type == 'min':
if min(h, w) < limit_side_len:
if h < w:
ratio = float(limit_side_len) / h
......@@ -182,6 +182,10 @@ class DetResizeForTest(object):
ratio = float(limit_side_len) / w
else:
ratio = 1.
elif self.limit_type == 'resize_long':
ratio = float(limit_side_len) / max(h,w)
else:
raise Exception('not support limit type, image ')
resize_h = int(h * ratio)
resize_w = int(w * ratio)
......
......@@ -44,16 +44,16 @@ class BaseRecLabelDecode(object):
self.character_str = string.printable[:-6]
dict_character = list(self.character_str)
elif character_type in support_character_type:
self.character_str = ""
self.character_str = []
assert character_dict_path is not None, "character_dict_path should not be None when character_type is {}".format(
character_type)
with open(character_dict_path, "rb") as fin:
lines = fin.readlines()
for line in lines:
line = line.decode('utf-8').strip("\n").strip("\r\n")
self.character_str += line
self.character_str.append(line)
if use_space_char:
self.character_str += " "
self.character_str.append(" ")
dict_character = list(self.character_str)
else:
......@@ -319,3 +319,156 @@ class SRNLabelDecode(BaseRecLabelDecode):
assert False, "unsupport type %s in get_beg_end_flag_idx" \
% beg_or_end
return idx
class TableLabelDecode(object):
""" """
def __init__(self,
max_text_length,
max_elem_length,
max_cell_num,
character_dict_path,
**kwargs):
self.max_text_length = max_text_length
self.max_elem_length = max_elem_length
self.max_cell_num = max_cell_num
list_character, list_elem = self.load_char_elem_dict(character_dict_path)
list_character = self.add_special_char(list_character)
list_elem = self.add_special_char(list_elem)
self.dict_character = {}
self.dict_idx_character = {}
for i, char in enumerate(list_character):
self.dict_idx_character[i] = char
self.dict_character[char] = i
self.dict_elem = {}
self.dict_idx_elem = {}
for i, elem in enumerate(list_elem):
self.dict_idx_elem[i] = elem
self.dict_elem[elem] = i
def load_char_elem_dict(self, character_dict_path):
list_character = []
list_elem = []
with open(character_dict_path, "rb") as fin:
lines = fin.readlines()
substr = lines[0].decode('utf-8').strip("\n").split("\t")
character_num = int(substr[0])
elem_num = int(substr[1])
for cno in range(1, 1 + character_num):
character = lines[cno].decode('utf-8').strip("\n")
list_character.append(character)
for eno in range(1 + character_num, 1 + character_num + elem_num):
elem = lines[eno].decode('utf-8').strip("\n")
list_elem.append(elem)
return list_character, list_elem
def add_special_char(self, list_character):
self.beg_str = "sos"
self.end_str = "eos"
list_character = [self.beg_str] + list_character + [self.end_str]
return list_character
def get_sp_tokens(self):
char_beg_idx = self.get_beg_end_flag_idx('beg', 'char')
char_end_idx = self.get_beg_end_flag_idx('end', 'char')
elem_beg_idx = self.get_beg_end_flag_idx('beg', 'elem')
elem_end_idx = self.get_beg_end_flag_idx('end', 'elem')
elem_char_idx1 = self.dict_elem['<td>']
elem_char_idx2 = self.dict_elem['<td']
sp_tokens = np.array([char_beg_idx, char_end_idx, elem_beg_idx,
elem_end_idx, elem_char_idx1, elem_char_idx2, self.max_text_length,
self.max_elem_length, self.max_cell_num])
return sp_tokens
def __call__(self, preds):
structure_probs = preds['structure_probs']
loc_preds = preds['loc_preds']
if isinstance(structure_probs,paddle.Tensor):
structure_probs = structure_probs.numpy()
if isinstance(loc_preds,paddle.Tensor):
loc_preds = loc_preds.numpy()
structure_idx = structure_probs.argmax(axis=2)
structure_probs = structure_probs.max(axis=2)
structure_str, structure_pos, result_score_list, result_elem_idx_list = self.decode(structure_idx,
structure_probs, 'elem')
res_html_code_list = []
res_loc_list = []
batch_num = len(structure_str)
for bno in range(batch_num):
res_loc = []
for sno in range(len(structure_str[bno])):
text = structure_str[bno][sno]
if text in ['<td>', '<td']:
pos = structure_pos[bno][sno]
res_loc.append(loc_preds[bno, pos])
res_html_code = ''.join(structure_str[bno])
res_loc = np.array(res_loc)
res_html_code_list.append(res_html_code)
res_loc_list.append(res_loc)
return {'res_html_code': res_html_code_list, 'res_loc': res_loc_list, 'res_score_list': result_score_list,
'res_elem_idx_list': result_elem_idx_list,'structure_str_list':structure_str}
def decode(self, text_index, structure_probs, char_or_elem):
"""convert text-label into text-index.
"""
if char_or_elem == "char":
current_dict = self.dict_idx_character
else:
current_dict = self.dict_idx_elem
ignored_tokens = self.get_ignored_tokens('elem')
beg_idx, end_idx = ignored_tokens
result_list = []
result_pos_list = []
result_score_list = []
result_elem_idx_list = []
batch_size = len(text_index)
for batch_idx in range(batch_size):
char_list = []
elem_pos_list = []
elem_idx_list = []
score_list = []
for idx in range(len(text_index[batch_idx])):
tmp_elem_idx = int(text_index[batch_idx][idx])
if idx > 0 and tmp_elem_idx == end_idx:
break
if tmp_elem_idx in ignored_tokens:
continue
char_list.append(current_dict[tmp_elem_idx])
elem_pos_list.append(idx)
score_list.append(structure_probs[batch_idx, idx])
elem_idx_list.append(tmp_elem_idx)
result_list.append(char_list)
result_pos_list.append(elem_pos_list)
result_score_list.append(score_list)
result_elem_idx_list.append(elem_idx_list)
return result_list, result_pos_list, result_score_list, result_elem_idx_list
def get_ignored_tokens(self, char_or_elem):
beg_idx = self.get_beg_end_flag_idx("beg", char_or_elem)
end_idx = self.get_beg_end_flag_idx("end", char_or_elem)
return [beg_idx, end_idx]
def get_beg_end_flag_idx(self, beg_or_end, char_or_elem):
if char_or_elem == "char":
if beg_or_end == "beg":
idx = self.dict_character[self.beg_str]
elif beg_or_end == "end":
idx = self.dict_character[self.end_str]
else:
assert False, "Unsupport type %s in get_beg_end_flag_idx of char" \
% beg_or_end
elif char_or_elem == "elem":
if beg_or_end == "beg":
idx = self.dict_elem[self.beg_str]
elif beg_or_end == "end":
idx = self.dict_elem[self.end_str]
else:
assert False, "Unsupport type %s in get_beg_end_flag_idx of elem" \
% beg_or_end
else:
assert False, "Unsupport type %s in char_or_elem" \
% char_or_elem
return idx
</overline>
α

$
ω
ψ
χ
(
υ
σ
,
ρ
ε
0
4
8
b
<
Ψ
Ω
D
3
Π
H
</strike>
L
Φ
Χ
θ
P
κ
λ
μ
T
ξ
X
β
γ
δ
\
ζ
η
`
d
<strike>
h
f
l
Θ
p
t
</sub>
x
Β
Γ
Δ
|
ǂ
ɛ
j
̧
̌
«
#
</b>
'
Ι
+
/
·
7
;
?
C
÷
G
K
<sup>
O
S
С
W
Α
[
_
c
z
g
<i>
o
<sub>
s
w
φ
ʹ
{
»
̆
e
ˆ
τ
ι
Ø
ß
×
˃
˂
"
i
&
π
*
æ
.
ø
Q
6
:
>
a
B
F
J
̄
N
R
V
<overline>
Z
^
¤
¥
§
<underline>
¢
£
­
Λ
©
n
r
°
±
v
<b>
k
~
̇
@
ł
®
!
</sup>
%
)
-
1
5
9
=
А
A
Σ
E
I
M
m
̨
</i>
U
Y
]
̸
2
̂
̀
́
̊
̈
q
u
ı
y
</underline>
̃
}
ν
此差异已折叠。
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import sys
import tarfile
import requests
from tqdm import tqdm
from ppocr.utils.logging import get_logger
def download_with_progressbar(url, save_path):
logger = get_logger()
response = requests.get(url, stream=True)
total_size_in_bytes = int(response.headers.get('content-length', 0))
block_size = 1024 # 1 Kibibyte
progress_bar = tqdm(total=total_size_in_bytes, unit='iB', unit_scale=True)
with open(save_path, 'wb') as file:
for data in response.iter_content(block_size):
progress_bar.update(len(data))
file.write(data)
progress_bar.close()
if total_size_in_bytes == 0 or progress_bar.n != total_size_in_bytes:
logger.error("Something went wrong while downloading models")
sys.exit(0)
def maybe_download(model_storage_directory, url):
# using custom model
tar_file_name_list = [
'inference.pdiparams', 'inference.pdiparams.info', 'inference.pdmodel'
]
if not os.path.exists(
os.path.join(model_storage_directory, 'inference.pdiparams')
) or not os.path.exists(
os.path.join(model_storage_directory, 'inference.pdmodel')):
assert url.endswith('.tar'), 'Only supports tar compressed package'
tmp_path = os.path.join(model_storage_directory, url.split('/')[-1])
print('download {} to {}'.format(url, tmp_path))
os.makedirs(model_storage_directory, exist_ok=True)
download_with_progressbar(url, tmp_path)
with tarfile.open(tmp_path, 'r') as tarObj:
for member in tarObj.getmembers():
filename = None
for tar_file_name in tar_file_name_list:
if tar_file_name in member.name:
filename = tar_file_name
if filename is None:
continue
file = tarObj.extractfile(member)
with open(
os.path.join(model_storage_directory, filename),
'wb') as f:
f.write(file.read())
os.remove(tmp_path)
def is_link(s):
return s is not None and s.startswith('http')
def confirm_model_dir_url(model_dir, default_model_dir, default_url):
url = default_url
if model_dir is None or is_link(model_dir):
if is_link(model_dir):
url = model_dir
file_name = url.split('/')[-1][:-4]
model_dir = default_model_dir
model_dir = os.path.join(model_dir, file_name)
return model_dir, url
include LICENSE
include README.md
recursive-include ppocr/utils *.txt utility.py logging.py network.py
recursive-include ppocr/data/ *.py
recursive-include ppocr/postprocess *.py
recursive-include tools/infer *.py
recursive-include ppstructure *.py
# PaddleStructure
## pipeline介绍
PaddleStructure 是一个用于复杂板式文字OCR的工具包,流程如下
![pipeline](../doc/table/pipeline.png)
在PaddleStructure中,图片会先经由layoutparser进行版面分析,在版面分析中,会对图片里的区域进行分类,根据根据类别进行对于的ocr流程。
目前layoutparser会输出五个类别:
1. Text
2. Title
3. Figure
4. List
5. Table
1-4类走传统的OCR流程,5走表格的OCR流程。
## LayoutParser
## Table OCR
[文档](table/README_ch.md)
## PaddleStructure whl包介绍
### 使用
1. 代码使用
```python
import cv2
from paddlestructure import PaddleStructure,draw_result
table_engine = PaddleStructure(
output='./output/table',
show_log=True)
img_path = '../doc/table/1.png'
img = cv2.imread(img_path)
result = table_engine(img)
for line in result:
print(line)
from PIL import Image
font_path = 'path/tp/PaddleOCR/doc/fonts/simfang.ttf'
image = Image.open(img_path).convert('RGB')
im_show = draw_result(image, result,font_path=font_path)
im_show = Image.fromarray(im_show)
im_show.save('result.jpg')
```
2. 命令行使用
```bash
paddlestructure --image_dir=../doc/table/1.png
```
### 参数说明
\ No newline at end of file
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from .paddlestructure import PaddleStructure, draw_result, to_excel
__all__ = ['PaddleStructure', 'draw_result', 'to_excel']
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import logging
import os
import sys
__dir__ = os.path.dirname(__file__)
sys.path.append(__dir__)
sys.path.append(os.path.join(__dir__, '..'))
import cv2
import numpy as np
from pathlib import Path
from ppocr.utils.logging import get_logger
from ppstructure.predict_system import OCRSystem, save_res
from ppstructure.table.predict_table import to_excel
from ppstructure.utility import init_args, draw_result
logger = get_logger()
from ppocr.utils.utility import check_and_read_gif, get_image_file_list
from ppocr.utils.network import maybe_download, download_with_progressbar, confirm_model_dir_url, is_link
__all__ = ['PaddleStructure', 'draw_result', 'to_excel']
VERSION = '2.1'
BASE_DIR = os.path.expanduser("~/.paddlestructure/")
model_urls = {
'det': 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_det_infer.tar',
'rec': 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_rec_infer.tar',
'structure': 'https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_structure_infer.tar'
}
def parse_args(mMain=True):
import argparse
parser = init_args()
parser.add_help = mMain
for action in parser._actions:
if action.dest in ['rec_char_dict_path', 'structure_char_dict_path']:
action.default = None
if mMain:
return parser.parse_args()
else:
inference_args_dict = {}
for action in parser._actions:
inference_args_dict[action.dest] = action.default
return argparse.Namespace(**inference_args_dict)
class PaddleStructure(OCRSystem):
def __init__(self, **kwargs):
params = parse_args(mMain=False)
params.__dict__.update(**kwargs)
if params.show_log:
logger.setLevel(logging.DEBUG)
params.use_angle_cls = False
# init model dir
params.det_model_dir, det_url = confirm_model_dir_url(params.det_model_dir,
os.path.join(BASE_DIR, VERSION, 'det'),
model_urls['det'])
params.rec_model_dir, rec_url = confirm_model_dir_url(params.rec_model_dir,
os.path.join(BASE_DIR, VERSION, 'rec'),
model_urls['rec'])
params.structure_model_dir, structure_url = confirm_model_dir_url(params.structure_model_dir,
os.path.join(BASE_DIR, VERSION, 'structure'),
model_urls['structure'])
# download model
maybe_download(params.det_model_dir, det_url)
maybe_download(params.rec_model_dir, rec_url)
maybe_download(params.structure_model_dir, structure_url)
if params.rec_char_dict_path is None:
params.rec_char_type = 'EN'
if os.path.exists(str(Path(__file__).parent / 'ppocr/utils/dict/table_dict.txt')):
params.rec_char_dict_path = str(Path(__file__).parent / 'ppocr/utils/dict/table_dict.txt')
else:
params.rec_char_dict_path = str(Path(__file__).parent.parent / 'ppocr/utils/dict/table_dict.txt')
if params.structure_char_dict_path is None:
if os.path.exists(str(Path(__file__).parent / 'ppocr/utils/dict/table_structure_dict.txt')):
params.structure_char_dict_path = str(
Path(__file__).parent / 'ppocr/utils/dict/table_structure_dict.txt')
else:
params.structure_char_dict_path = str(
Path(__file__).parent.parent / 'ppocr/utils/dict/table_structure_dict.txt')
print(params)
super().__init__(params)
def __call__(self, img):
if isinstance(img, str):
# download net image
if img.startswith('http'):
download_with_progressbar(img, 'tmp.jpg')
img = 'tmp.jpg'
image_file = img
img, flag = check_and_read_gif(image_file)
if not flag:
with open(image_file, 'rb') as f:
np_arr = np.frombuffer(f.read(), dtype=np.uint8)
img = cv2.imdecode(np_arr, cv2.IMREAD_COLOR)
if img is None:
logger.error("error in loading image:{}".format(image_file))
return None
if isinstance(img, np.ndarray) and len(img.shape) == 2:
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
res = super().__call__(img)
return res
def main():
# for cmd
args = parse_args(mMain=True)
image_dir = args.image_dir
save_folder = args.output
if image_dir.startswith('http'):
download_with_progressbar(image_dir, 'tmp.jpg')
image_file_list = ['tmp.jpg']
else:
image_file_list = get_image_file_list(args.image_dir)
if len(image_file_list) == 0:
logger.error('no images find in {}'.format(args.image_dir))
return
structure_engine = PaddleStructure(**(args.__dict__))
for img_path in image_file_list:
img_name = os.path.basename(img_path).split('.')[0]
logger.info('{}{}{}'.format('*' * 10, img_path, '*' * 10))
result = structure_engine(img_path)
for item in result:
logger.info(item['res'])
save_res(result, save_folder, img_name)
logger.info('result save to {}'.format(os.path.join(save_folder, img_name)))
\ No newline at end of file
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import sys
import subprocess
__dir__ = os.path.dirname(os.path.abspath(__file__))
sys.path.append(__dir__)
sys.path.append(os.path.abspath(os.path.join(__dir__, '..')))
os.environ["FLAGS_allocator_strategy"] = 'auto_growth'
import cv2
import numpy as np
import time
import layoutparser as lp
from ppocr.utils.utility import get_image_file_list, check_and_read_gif
from ppocr.utils.logging import get_logger
from tools.infer.predict_system import TextSystem
from ppstructure.table.predict_table import TableSystem, to_excel
from ppstructure.utility import parse_args,draw_result
logger = get_logger()
class OCRSystem(object):
def __init__(self, args):
args.det_limit_type = 'resize_long'
args.drop_score = 0
self.text_system = TextSystem(args)
self.table_system = TableSystem(args, self.text_system.text_detector, self.text_system.text_recognizer)
self.table_layout = lp.PaddleDetectionLayoutModel("lp://PubLayNet/ppyolov2_r50vd_dcn_365e_publaynet/config",
threshold=0.5, enable_mkldnn=args.enable_mkldnn,
enforce_cpu=not args.use_gpu, thread_num=args.cpu_threads)
self.use_angle_cls = args.use_angle_cls
self.drop_score = args.drop_score
def __call__(self, img):
ori_im = img.copy()
layout_res = self.table_layout.detect(img[..., ::-1])
res_list = []
for region in layout_res:
x1, y1, x2, y2 = region.coordinates
x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
roi_img = ori_im[y1:y2, x1:x2, :]
if region.type == 'Table':
res = self.table_system(roi_img)
else:
filter_boxes, filter_rec_res = self.text_system(roi_img)
filter_boxes = [x + [x1, y1] for x in filter_boxes]
filter_boxes = [x.reshape(-1).tolist() for x in filter_boxes]
res = (filter_boxes, filter_rec_res)
res_list.append({'type': region.type, 'bbox': [x1, y1, x2, y2], 'res': res})
return res_list
def save_res(res, save_folder, img_name):
excel_save_folder = os.path.join(save_folder, img_name)
os.makedirs(excel_save_folder, exist_ok=True)
# save res
for region in res:
if region['type'] == 'Table':
excel_path = os.path.join(excel_save_folder, '{}.xlsx'.format(region['bbox']))
to_excel(region['res'], excel_path)
elif region['type'] == 'Figure':
pass
else:
with open(os.path.join(excel_save_folder, 'res.txt'), 'a', encoding='utf8') as f:
for box, rec_res in zip(region['res'][0], region['res'][1]):
f.write('{}\t{}\n'.format(np.array(box).reshape(-1).tolist(), rec_res))
def main(args):
image_file_list = get_image_file_list(args.image_dir)
image_file_list = image_file_list
image_file_list = image_file_list[args.process_id::args.total_process_num]
save_folder = args.output
os.makedirs(save_folder, exist_ok=True)
structure_sys = OCRSystem(args)
img_num = len(image_file_list)
for i, image_file in enumerate(image_file_list):
logger.info("[{}/{}] {}".format(i, img_num, image_file))
img, flag = check_and_read_gif(image_file)
img_name = os.path.basename(image_file).split('.')[0]
if not flag:
img = cv2.imread(image_file)
if img is None:
logger.error("error in loading image:{}".format(image_file))
continue
starttime = time.time()
res = structure_sys(img)
save_res(res, save_folder, img_name)
draw_img = draw_result(img,res, args.vis_font_path)
cv2.imwrite(os.path.join(save_folder, img_name, 'show.jpg'), draw_img)
logger.info('result save to {}'.format(os.path.join(save_folder, img_name)))
elapse = time.time() - starttime
logger.info("Predict time : {:.3f}s".format(elapse))
if __name__ == "__main__":
args = parse_args()
if args.use_mp:
p_list = []
total_process_num = args.total_process_num
for process_id in range(total_process_num):
cmd = [sys.executable, "-u"] + sys.argv + [
"--process_id={}".format(process_id),
"--use_mp={}".format(False)
]
p = subprocess.Popen(cmd, stdout=sys.stdout, stderr=sys.stdout)
p_list.append(p)
for p in p_list:
p.wait()
else:
main(args)
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
from setuptools import setup
from io import open
import shutil
with open('../requirements.txt', encoding="utf-8-sig") as f:
requirements = f.readlines()
requirements.append('tqdm')
requirements.append('layoutparser')
requirements.append('iopath')
def readme():
with open('README_ch.md', encoding="utf-8-sig") as f:
README = f.read()
return README
shutil.copytree('../ppstructure/table', './ppstructure/table')
shutil.copyfile('../ppstructure/predict_system.py', './ppstructure/predict_system.py')
shutil.copyfile('../ppstructure/utility.py', './ppstructure/utility.py')
shutil.copytree('../ppocr', './ppocr')
shutil.copytree('../tools', './tools')
shutil.copyfile('../LICENSE', './LICENSE')
setup(
name='paddlestructure',
packages=['paddlestructure'],
package_dir={'paddlestructure': ''},
include_package_data=True,
entry_points={"console_scripts": ["paddlestructure= paddlestructure.paddlestructure:main"]},
version='1.0',
install_requires=requirements,
license='Apache License 2.0',
description='Awesome OCR toolkits based on PaddlePaddle (8.6M ultra-lightweight pre-trained model, support training and deployment among server, mobile, embeded and IoT devices',
long_description=readme(),
long_description_content_type='text/markdown',
url='https://github.com/PaddlePaddle/PaddleOCR',
download_url='https://github.com/PaddlePaddle/PaddleOCR.git',
keywords=[
'ocr textdetection textrecognition paddleocr crnn east star-net rosetta ocrlite db chineseocr chinesetextdetection chinesetextrecognition'
],
classifiers=[
'Intended Audience :: Developers', 'Operating System :: OS Independent',
'Natural Language :: Chinese (Simplified)',
'Programming Language :: Python :: 3',
'Programming Language :: Python :: 3.2',
'Programming Language :: Python :: 3.3',
'Programming Language :: Python :: 3.4',
'Programming Language :: Python :: 3.5',
'Programming Language :: Python :: 3.6',
'Programming Language :: Python :: 3.7', 'Topic :: Utilities'
], )
shutil.rmtree('ppocr')
shutil.rmtree('tools')
shutil.rmtree('ppstructure')
os.remove('LICENSE')
# 表格结构和内容预测
## pipeline
表格的ocr主要包含三个模型
1. 单行文本检测-DB
2. 单行文本识别-CRNN
3. 表格结构和cell坐标预测-RARE
具体流程图如下
![tableocr_pipeline](../../doc/table/tableocr_pipeline.png)
1. 图片由单行文字检测检测到单行文字的坐标,然后送入识别模型拿到识别结果。
2. 图片由表格结构和cell坐标预测拿到表格的结构信息和单元格的坐标信息。
3. 由单行文字的坐标、识别结果和单元格的坐标一起组合出单元格的识别结果。
4. 单元格的识别结果和表格结构一起构造表格的html字符串。
## 使用
### 训练
TBD
### 评估
先cd到PaddleOCR/ppstructure目录下
表格使用 TEDS(Tree-Edit-Distance-based Similarity) 作为模型的评估指标。在进行模型评估之前,需要将pipeline中的三个模型分别导出为inference模型(我们已经提供好),还需要准备评估的gt, gt示例如下:
```json
{"PMC4289340_004_00.png": [["<html>", "<body>", "<table>", "<thead>", "<tr>", "<td>", "</td>", "<td>", "</td>", "<td>", "</td>", "</tr>", "</thead>", "<tbody>", "<tr>", "<td>", "</td>", "<td>", "</td>", "<td>", "</td>", "</tr>", "</tbody>", "</table>", "</body>", "</html>"], [[1, 4, 29, 13], [137, 4, 161, 13], [215, 4, 236, 13], [1, 17, 30, 27], [137, 17, 147, 27], [215, 17, 225, 27]], [["<b>", "F", "e", "a", "t", "u", "r", "e", "</b>"], ["<b>", "G", "b", "3", " ", "+", "</b>"], ["<b>", "G", "b", "3", " ", "-", "</b>"], ["<b>", "P", "a", "t", "i", "e", "n", "t", "s", "</b>"], ["6", "2"], ["4", "5"]]]}
```
示例对应的表格如下
![tableocr_pipeline](../../doc/table/table_example.png)
准备完成后使用如下命令进行评估,评估完成后会输出teds指标。
```python
python3 table/eval_table.py --det_model_dir=path/to/det_model_dir --rec_model_dir=path/to/rec_model_dir --structure_model_dir=path/to/structure_model_dir --image_dir=../doc/table/1.png --rec_char_dict_path=../ppocr/utils/dict/table_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --rec_char_type=EN --det_limit_side_len=736 --det_limit_type=min --gt_path=path/to/gt.json
```
### 预测
先cd到PaddleOCR/ppstructure目录下
```python
python3 table/predict_table.py --det_model_dir=path/to/det_model_dir --rec_model_dir=path/to/rec_model_dir --structure_model_dir=path/to/structure_model_dir --image_dir=../doc/table/1.png --rec_char_dict_path=../ppocr/utils/dict/table_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --rec_char_type=EN --det_limit_side_len=736 --det_limit_type=min --output ../output/table
```
运行完成后,每张图片的excel表格会保存到table_output字段指定的目录下
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import sys
__dir__ = os.path.dirname(os.path.abspath(__file__))
sys.path.append(__dir__)
sys.path.append(os.path.abspath(os.path.join(__dir__, '../..')))
import cv2
import json
from tqdm import tqdm
from ppstructure.table.table_metric import TEDS
from ppstructure.table.predict_table import TableSystem
from ppstructure.utility import init_args
from ppocr.utils.logging import get_logger
logger = get_logger()
def parse_args():
parser = init_args()
parser.add_argument("--gt_path", type=str)
return parser.parse_args()
def main(gt_path, img_root, args):
teds = TEDS(n_jobs=16)
text_sys = TableSystem(args)
jsons_gt = json.load(open(gt_path)) # gt
pred_htmls = []
gt_htmls = []
for img_name in tqdm(jsons_gt):
# read image
img = cv2.imread(os.path.join(img_root,img_name))
pred_html = text_sys(img)
pred_htmls.append(pred_html)
gt_structures, gt_bboxes, gt_contents, contents_with_block = jsons_gt[img_name]
gt_html, gt = get_gt_html(gt_structures, contents_with_block)
gt_htmls.append(gt_html)
scores = teds.batch_evaluate_html(gt_htmls, pred_htmls)
logger.info('teds:', sum(scores) / len(scores))
def get_gt_html(gt_structures, contents_with_block):
end_html = []
td_index = 0
for tag in gt_structures:
if '</td>' in tag:
if contents_with_block[td_index] != []:
end_html.extend(contents_with_block[td_index])
end_html.append(tag)
td_index += 1
else:
end_html.append(tag)
return ''.join(end_html), end_html
if __name__ == '__main__':
args = parse_args()
main(args.gt_path,args.image_dir, args)
import json
def distance(box_1, box_2):
x1, y1, x2, y2 = box_1
x3, y3, x4, y4 = box_2
dis = abs(x3 - x1) + abs(y3 - y1) + abs(x4- x2) + abs(y4 - y2)
dis_2 = abs(x3 - x1) + abs(y3 - y1)
dis_3 = abs(x4- x2) + abs(y4 - y2)
return dis + min(dis_2, dis_3)
def compute_iou(rec1, rec2):
"""
computing IoU
:param rec1: (y0, x0, y1, x1), which reflects
(top, left, bottom, right)
:param rec2: (y0, x0, y1, x1)
:return: scala value of IoU
"""
# computing area of each rectangles
S_rec1 = (rec1[2] - rec1[0]) * (rec1[3] - rec1[1])
S_rec2 = (rec2[2] - rec2[0]) * (rec2[3] - rec2[1])
# computing the sum_area
sum_area = S_rec1 + S_rec2
# find the each edge of intersect rectangle
left_line = max(rec1[1], rec2[1])
right_line = min(rec1[3], rec2[3])
top_line = max(rec1[0], rec2[0])
bottom_line = min(rec1[2], rec2[2])
# judge if there is an intersect
if left_line >= right_line or top_line >= bottom_line:
return 0.0
else:
intersect = (right_line - left_line) * (bottom_line - top_line)
return (intersect / (sum_area - intersect))*1.0
def matcher_merge(ocr_bboxes, pred_bboxes):
all_dis = []
ious = []
matched = {}
for i, gt_box in enumerate(ocr_bboxes):
distances = []
for j, pred_box in enumerate(pred_bboxes):
# compute l1 distence and IOU between two boxes
distances.append((distance(gt_box, pred_box), 1. - compute_iou(gt_box, pred_box)))
sorted_distances = distances.copy()
# select nearest cell
sorted_distances = sorted(sorted_distances, key = lambda item: (item[1], item[0]))
if distances.index(sorted_distances[0]) not in matched.keys():
matched[distances.index(sorted_distances[0])] = [i]
else:
matched[distances.index(sorted_distances[0])].append(i)
return matched#, sum(ious) / len(ious)
def complex_num(pred_bboxes):
complex_nums = []
for bbox in pred_bboxes:
distances = []
temp_ious = []
for pred_bbox in pred_bboxes:
if bbox != pred_bbox:
distances.append(distance(bbox, pred_bbox))
temp_ious.append(compute_iou(bbox, pred_bbox))
complex_nums.append(temp_ious[distances.index(min(distances))])
return sum(complex_nums) / len(complex_nums)
def get_rows(pred_bboxes):
pre_bbox = pred_bboxes[0]
res = []
step = 0
for i in range(len(pred_bboxes)):
bbox = pred_bboxes[i]
if bbox[1] - pre_bbox[1] > 2 or bbox[0] - pre_bbox[0] < 0:
break
else:
res.append(bbox)
step += 1
for i in range(step):
pred_bboxes.pop(0)
return res, pred_bboxes
def refine_rows(pred_bboxes): # 微调整行的框,使在一条水平线上
ys_1 = []
ys_2 = []
for box in pred_bboxes:
ys_1.append(box[1])
ys_2.append(box[3])
min_y_1 = sum(ys_1) / len(ys_1)
min_y_2 = sum(ys_2) / len(ys_2)
re_boxes = []
for box in pred_bboxes:
box[1] = min_y_1
box[3] = min_y_2
re_boxes.append(box)
return re_boxes
def matcher_refine_row(gt_bboxes, pred_bboxes):
before_refine_pred_bboxes = pred_bboxes.copy()
pred_bboxes = []
while(len(before_refine_pred_bboxes) != 0):
row_bboxes, before_refine_pred_bboxes = get_rows(before_refine_pred_bboxes)
print(row_bboxes)
pred_bboxes.extend(refine_rows(row_bboxes))
all_dis = []
ious = []
matched = {}
for i, gt_box in enumerate(gt_bboxes):
distances = []
#temp_ious = []
for j, pred_box in enumerate(pred_bboxes):
distances.append(distance(gt_box, pred_box))
#temp_ious.append(compute_iou(gt_box, pred_box))
#all_dis.append(min(distances))
#ious.append(temp_ious[distances.index(min(distances))])
if distances.index(min(distances)) not in matched.keys():
matched[distances.index(min(distances))] = [i]
else:
matched[distances.index(min(distances))].append(i)
return matched#, sum(ious) / len(ious)
#先挑选出一行,再进行匹配
def matcher_structure_1(gt_bboxes, pred_bboxes_rows, pred_bboxes):
gt_box_index = 0
delete_gt_bboxes = gt_bboxes.copy()
match_bboxes_ready = []
matched = {}
while(len(delete_gt_bboxes) != 0):
row_bboxes, delete_gt_bboxes = get_rows(delete_gt_bboxes)
row_bboxes = sorted(row_bboxes, key = lambda key: key[0])
if len(pred_bboxes_rows) > 0:
match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
print(row_bboxes)
for i, gt_box in enumerate(row_bboxes):
#print(gt_box)
pred_distances = []
distances = []
for pred_bbox in pred_bboxes:
pred_distances.append(distance(gt_box, pred_bbox))
for j, pred_box in enumerate(match_bboxes_ready):
distances.append(distance(gt_box, pred_box))
index = pred_distances.index(min(distances))
#print('index', index)
if index not in matched.keys():
matched[index] = [gt_box_index]
else:
matched[index].append(gt_box_index)
gt_box_index += 1
return matched
def matcher_structure(gt_bboxes, pred_bboxes_rows, pred_bboxes):
'''
gt_bboxes: 排序后
pred_bboxes:
'''
pre_bbox = gt_bboxes[0]
matched = {}
match_bboxes_ready = []
match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
for i, gt_box in enumerate(gt_bboxes):
pred_distances = []
for pred_bbox in pred_bboxes:
pred_distances.append(distance(gt_box, pred_bbox))
distances = []
gap_pre = gt_box[1] - pre_bbox[1]
gap_pre_1 = gt_box[0] - pre_bbox[2]
#print(gap_pre, len(pred_bboxes_rows))
if (gap_pre_1 < 0 and len(pred_bboxes_rows) > 0):
match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
if len(pred_bboxes_rows) == 1:
match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
if len(match_bboxes_ready) == 0 and len(pred_bboxes_rows) > 0:
match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
if len(match_bboxes_ready) == 0 and len(pred_bboxes_rows) == 0:
break
#print(match_bboxes_ready)
for j, pred_box in enumerate(match_bboxes_ready):
distances.append(distance(gt_box, pred_box))
index = pred_distances.index(min(distances))
#print(gt_box, index)
#match_bboxes_ready.pop(distances.index(min(distances)))
print(gt_box, match_bboxes_ready[distances.index(min(distances))])
if index not in matched.keys():
matched[index] = [i]
else:
matched[index].append(i)
pre_bbox = gt_box
return matched
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import sys
__dir__ = os.path.dirname(os.path.abspath(__file__))
sys.path.append(__dir__)
sys.path.append(os.path.abspath(os.path.join(__dir__, '../..')))
os.environ["FLAGS_allocator_strategy"] = 'auto_growth'
import cv2
import numpy as np
import math
import time
import traceback
import paddle
import tools.infer.utility as utility
from ppocr.data import create_operators, transform
from ppocr.postprocess import build_post_process
from ppocr.utils.logging import get_logger
from ppocr.utils.utility import get_image_file_list, check_and_read_gif
logger = get_logger()
class TableStructurer(object):
def __init__(self, args):
pre_process_list = [{
'ResizeTableImage': {
'max_len': args.structure_max_len
}
}, {
'NormalizeImage': {
'std': [0.229, 0.224, 0.225],
'mean': [0.485, 0.456, 0.406],
'scale': '1./255.',
'order': 'hwc'
}
}, {
'PaddingTableImage': None
}, {
'ToCHWImage': None
}, {
'KeepKeys': {
'keep_keys': ['image']
}
}]
postprocess_params = {
'name': 'TableLabelDecode',
"character_type": args.structure_char_type,
"character_dict_path": args.structure_char_dict_path,
"max_text_length": args.structure_max_text_length,
"max_elem_length": args.structure_max_elem_length,
"max_cell_num": args.structure_max_cell_num
}
self.preprocess_op = create_operators(pre_process_list)
self.postprocess_op = build_post_process(postprocess_params)
self.predictor, self.input_tensor, self.output_tensors = \
utility.create_predictor(args, 'structure', logger)
def __call__(self, img):
ori_im = img.copy()
data = {'image': img}
data = transform(data, self.preprocess_op)
img = data[0]
if img is None:
return None, 0
img = np.expand_dims(img, axis=0)
img = img.copy()
starttime = time.time()
self.input_tensor.copy_from_cpu(img)
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
preds = {}
preds['structure_probs'] = outputs[1]
preds['loc_preds'] = outputs[0]
post_result = self.postprocess_op(preds)
structure_str_list = post_result['structure_str_list']
res_loc = post_result['res_loc']
imgh, imgw = ori_im.shape[0:2]
res_loc_final = []
for rno in range(len(res_loc[0])):
x0, y0, x1, y1 = res_loc[0][rno]
left = max(int(imgw * x0), 0)
top = max(int(imgh * y0), 0)
right = min(int(imgw * x1), imgw - 1)
bottom = min(int(imgh * y1), imgh - 1)
res_loc_final.append([left, top, right, bottom])
structure_str_list = structure_str_list[0][:-1]
structure_str_list = ['<html>', '<body>', '<table>'] + structure_str_list + ['</table>', '</body>', '</html>']
elapse = time.time() - starttime
return (structure_str_list, res_loc_final), elapse
def main(args):
image_file_list = get_image_file_list(args.image_dir)
table_structurer = TableStructurer(args)
count = 0
total_time = 0
for image_file in image_file_list:
img, flag = check_and_read_gif(image_file)
if not flag:
img = cv2.imread(image_file)
if img is None:
logger.info("error in loading image:{}".format(image_file))
continue
structure_res, elapse = table_structurer(img)
logger.info("result: {}".format(structure_res))
if count > 0:
total_time += elapse
count += 1
logger.info("Predict time of {}: {}".format(image_file, elapse))
if __name__ == "__main__":
main(utility.parse_args())
# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import sys
import subprocess
__dir__ = os.path.dirname(os.path.abspath(__file__))
sys.path.append(__dir__)
sys.path.append(os.path.abspath(os.path.join(__dir__, '..')))
sys.path.append(os.path.abspath(os.path.join(__dir__, '../..')))
os.environ["FLAGS_allocator_strategy"] = 'auto_growth'
import cv2
import copy
import numpy as np
import time
import tools.infer.predict_rec as predict_rec
import tools.infer.predict_det as predict_det
from ppocr.utils.utility import get_image_file_list, check_and_read_gif
from ppocr.utils.logging import get_logger
from ppstructure.table.matcher import distance, compute_iou
from ppstructure.utility import parse_args
import ppstructure.table.predict_structure as predict_strture
logger = get_logger()
def expand(pix, det_box, shape):
x0, y0, x1, y1 = det_box
# print(shape)
h, w, c = shape
tmp_x0 = x0 - pix
tmp_x1 = x1 + pix
tmp_y0 = y0 - pix
tmp_y1 = y1 + pix
x0_ = tmp_x0 if tmp_x0 >= 0 else 0
x1_ = tmp_x1 if tmp_x1 <= w else w
y0_ = tmp_y0 if tmp_y0 >= 0 else 0
y1_ = tmp_y1 if tmp_y1 <= h else h
return x0_, y0_, x1_, y1_
class TableSystem(object):
def __init__(self, args, text_detector=None, text_recognizer=None):
self.text_detector = predict_det.TextDetector(args) if text_detector is None else text_detector
self.text_recognizer = predict_rec.TextRecognizer(args) if text_recognizer is None else text_recognizer
self.table_structurer = predict_strture.TableStructurer(args)
def __call__(self, img):
ori_im = img.copy()
structure_res, elapse = self.table_structurer(copy.deepcopy(img))
dt_boxes, elapse = self.text_detector(copy.deepcopy(img))
dt_boxes = sorted_boxes(dt_boxes)
r_boxes = []
for box in dt_boxes:
x_min = box[:, 0].min() - 1
x_max = box[:, 0].max() + 1
y_min = box[:, 1].min() - 1
y_max = box[:, 1].max() + 1
box = [x_min, y_min, x_max, y_max]
r_boxes.append(box)
dt_boxes = np.array(r_boxes)
logger.debug("dt_boxes num : {}, elapse : {}".format(
len(dt_boxes), elapse))
if dt_boxes is None:
return None, None
img_crop_list = []
for i in range(len(dt_boxes)):
det_box = dt_boxes[i]
x0, y0, x1, y1 = expand(2, det_box, ori_im.shape)
text_rect = ori_im[int(y0):int(y1), int(x0):int(x1), :]
img_crop_list.append(text_rect)
rec_res, elapse = self.text_recognizer(img_crop_list)
logger.debug("rec_res num : {}, elapse : {}".format(
len(rec_res), elapse))
pred_html, pred = self.rebuild_table(structure_res, dt_boxes, rec_res)
return pred_html
def rebuild_table(self, structure_res, dt_boxes, rec_res):
pred_structures, pred_bboxes = structure_res
matched_index = self.match_result(dt_boxes, pred_bboxes)
pred_html, pred = self.get_pred_html(pred_structures, matched_index, rec_res)
return pred_html, pred
def match_result(self, dt_boxes, pred_bboxes):
matched = {}
for i, gt_box in enumerate(dt_boxes):
# gt_box = [np.min(gt_box[:, 0]), np.min(gt_box[:, 1]), np.max(gt_box[:, 0]), np.max(gt_box[:, 1])]
distances = []
for j, pred_box in enumerate(pred_bboxes):
distances.append(
(distance(gt_box, pred_box), 1. - compute_iou(gt_box, pred_box))) # 获取两两cell之间的L1距离和 1- IOU
sorted_distances = distances.copy()
# 根据距离和IOU挑选最"近"的cell
sorted_distances = sorted(sorted_distances, key=lambda item: (item[1], item[0]))
if distances.index(sorted_distances[0]) not in matched.keys():
matched[distances.index(sorted_distances[0])] = [i]
else:
matched[distances.index(sorted_distances[0])].append(i)
return matched
def get_pred_html(self, pred_structures, matched_index, ocr_contents):
end_html = []
td_index = 0
for tag in pred_structures:
if '</td>' in tag:
if td_index in matched_index.keys():
b_with = False
if '<b>' in ocr_contents[matched_index[td_index][0]] and len(matched_index[td_index]) > 1:
b_with = True
end_html.extend('<b>')
for i, td_index_index in enumerate(matched_index[td_index]):
content = ocr_contents[td_index_index][0]
if len(matched_index[td_index]) > 1:
if len(content) == 0:
continue
if content[0] == ' ':
content = content[1:]
if '<b>' in content:
content = content[3:]
if '</b>' in content:
content = content[:-4]
if len(content) == 0:
continue
if i != len(matched_index[td_index]) - 1 and ' ' != content[-1]:
content += ' '
end_html.extend(content)
if b_with:
end_html.extend('</b>')
end_html.append(tag)
td_index += 1
else:
end_html.append(tag)
return ''.join(end_html), end_html
def sorted_boxes(dt_boxes):
"""
Sort text boxes in order from top to bottom, left to right
args:
dt_boxes(array):detected text boxes with shape [4, 2]
return:
sorted boxes(array) with shape [4, 2]
"""
num_boxes = dt_boxes.shape[0]
sorted_boxes = sorted(dt_boxes, key=lambda x: (x[0][1], x[0][0]))
_boxes = list(sorted_boxes)
for i in range(num_boxes - 1):
if abs(_boxes[i + 1][0][1] - _boxes[i][0][1]) < 10 and \
(_boxes[i + 1][0][0] < _boxes[i][0][0]):
tmp = _boxes[i]
_boxes[i] = _boxes[i + 1]
_boxes[i + 1] = tmp
return _boxes
def to_excel(html_table, excel_path):
from tablepyxl import tablepyxl
tablepyxl.document_to_xl(html_table, excel_path)
def main(args):
image_file_list = get_image_file_list(args.image_dir)
image_file_list = image_file_list[args.process_id::args.total_process_num]
os.makedirs(args.output, exist_ok=True)
text_sys = TableSystem(args)
img_num = len(image_file_list)
for i, image_file in enumerate(image_file_list):
logger.info("[{}/{}] {}".format(i, img_num, image_file))
img, flag = check_and_read_gif(image_file)
excel_path = os.path.join(args.table_output, os.path.basename(image_file).split('.')[0] + '.xlsx')
if not flag:
img = cv2.imread(image_file)
if img is None:
logger.error("error in loading image:{}".format(image_file))
continue
starttime = time.time()
pred_html = text_sys(img)
to_excel(pred_html, excel_path)
logger.info('excel saved to {}'.format(excel_path))
logger.info(pred_html)
elapse = time.time() - starttime
logger.info("Predict time : {:.3f}s".format(elapse))
if __name__ == "__main__":
args = parse_args()
if args.use_mp:
p_list = []
total_process_num = args.total_process_num
for process_id in range(total_process_num):
cmd = [sys.executable, "-u"] + sys.argv + [
"--process_id={}".format(process_id),
"--use_mp={}".format(False)
]
p = subprocess.Popen(cmd, stdout=sys.stdout, stderr=sys.stdout)
p_list.append(p)
for p in p_list:
p.wait()
else:
main(args)
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
__all__ = ['TEDS']
from .table_metric import TEDS
\ No newline at end of file
from tqdm import tqdm
from concurrent.futures import ProcessPoolExecutor, as_completed
def parallel_process(array, function, n_jobs=16, use_kwargs=False, front_num=0):
"""
A parallel version of the map function with a progress bar.
Args:
array (array-like): An array to iterate over.
function (function): A python function to apply to the elements of array
n_jobs (int, default=16): The number of cores to use
use_kwargs (boolean, default=False): Whether to consider the elements of array as dictionaries of
keyword arguments to function
front_num (int, default=3): The number of iterations to run serially before kicking off the parallel job.
Useful for catching bugs
Returns:
[function(array[0]), function(array[1]), ...]
"""
# We run the first few iterations serially to catch bugs
if front_num > 0:
front = [function(**a) if use_kwargs else function(a)
for a in array[:front_num]]
else:
front = []
# If we set n_jobs to 1, just run a list comprehension. This is useful for benchmarking and debugging.
if n_jobs == 1:
return front + [function(**a) if use_kwargs else function(a) for a in tqdm(array[front_num:])]
# Assemble the workers
with ProcessPoolExecutor(max_workers=n_jobs) as pool:
# Pass the elements of array into function
if use_kwargs:
futures = [pool.submit(function, **a) for a in array[front_num:]]
else:
futures = [pool.submit(function, a) for a in array[front_num:]]
kwargs = {
'total': len(futures),
'unit': 'it',
'unit_scale': True,
'leave': True
}
# Print out the progress as tasks complete
for f in tqdm(as_completed(futures), **kwargs):
pass
out = []
# Get the results from the futures.
for i, future in tqdm(enumerate(futures)):
try:
out.append(future.result())
except Exception as e:
out.append(e)
return front + out
# Copyright 2020 IBM
# Author: peter.zhong@au1.ibm.com
#
# This is free software; you can redistribute it and/or modify
# it under the terms of the Apache 2.0 License.
#
# This software is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# Apache 2.0 License for more details.
import distance
from apted import APTED, Config
from apted.helpers import Tree
from lxml import etree, html
from collections import deque
from .parallel import parallel_process
from tqdm import tqdm
class TableTree(Tree):
def __init__(self, tag, colspan=None, rowspan=None, content=None, *children):
self.tag = tag
self.colspan = colspan
self.rowspan = rowspan
self.content = content
self.children = list(children)
def bracket(self):
"""Show tree using brackets notation"""
if self.tag == 'td':
result = '"tag": %s, "colspan": %d, "rowspan": %d, "text": %s' % \
(self.tag, self.colspan, self.rowspan, self.content)
else:
result = '"tag": %s' % self.tag
for child in self.children:
result += child.bracket()
return "{{{}}}".format(result)
class CustomConfig(Config):
@staticmethod
def maximum(*sequences):
"""Get maximum possible value
"""
return max(map(len, sequences))
def normalized_distance(self, *sequences):
"""Get distance from 0 to 1
"""
return float(distance.levenshtein(*sequences)) / self.maximum(*sequences)
def rename(self, node1, node2):
"""Compares attributes of trees"""
#print(node1.tag)
if (node1.tag != node2.tag) or (node1.colspan != node2.colspan) or (node1.rowspan != node2.rowspan):
return 1.
if node1.tag == 'td':
if node1.content or node2.content:
#print(node1.content, )
return self.normalized_distance(node1.content, node2.content)
return 0.
class CustomConfig_del_short(Config):
@staticmethod
def maximum(*sequences):
"""Get maximum possible value
"""
return max(map(len, sequences))
def normalized_distance(self, *sequences):
"""Get distance from 0 to 1
"""
return float(distance.levenshtein(*sequences)) / self.maximum(*sequences)
def rename(self, node1, node2):
"""Compares attributes of trees"""
if (node1.tag != node2.tag) or (node1.colspan != node2.colspan) or (node1.rowspan != node2.rowspan):
return 1.
if node1.tag == 'td':
if node1.content or node2.content:
#print('before')
#print(node1.content, node2.content)
#print('after')
node1_content = node1.content
node2_content = node2.content
if len(node1_content) < 3:
node1_content = ['####']
if len(node2_content) < 3:
node2_content = ['####']
return self.normalized_distance(node1_content, node2_content)
return 0.
class CustomConfig_del_block(Config):
@staticmethod
def maximum(*sequences):
"""Get maximum possible value
"""
return max(map(len, sequences))
def normalized_distance(self, *sequences):
"""Get distance from 0 to 1
"""
return float(distance.levenshtein(*sequences)) / self.maximum(*sequences)
def rename(self, node1, node2):
"""Compares attributes of trees"""
if (node1.tag != node2.tag) or (node1.colspan != node2.colspan) or (node1.rowspan != node2.rowspan):
return 1.
if node1.tag == 'td':
if node1.content or node2.content:
node1_content = node1.content
node2_content = node2.content
while ' ' in node1_content:
print(node1_content.index(' '))
node1_content.pop(node1_content.index(' '))
while ' ' in node2_content:
print(node2_content.index(' '))
node2_content.pop(node2_content.index(' '))
return self.normalized_distance(node1_content, node2_content)
return 0.
class TEDS(object):
''' Tree Edit Distance basead Similarity
'''
def __init__(self, structure_only=False, n_jobs=1, ignore_nodes=None):
assert isinstance(n_jobs, int) and (
n_jobs >= 1), 'n_jobs must be an integer greather than 1'
self.structure_only = structure_only
self.n_jobs = n_jobs
self.ignore_nodes = ignore_nodes
self.__tokens__ = []
def tokenize(self, node):
''' Tokenizes table cells
'''
self.__tokens__.append('<%s>' % node.tag)
if node.text is not None:
self.__tokens__ += list(node.text)
for n in node.getchildren():
self.tokenize(n)
if node.tag != 'unk':
self.__tokens__.append('</%s>' % node.tag)
if node.tag != 'td' and node.tail is not None:
self.__tokens__ += list(node.tail)
def load_html_tree(self, node, parent=None):
''' Converts HTML tree to the format required by apted
'''
global __tokens__
if node.tag == 'td':
if self.structure_only:
cell = []
else:
self.__tokens__ = []
self.tokenize(node)
cell = self.__tokens__[1:-1].copy()
new_node = TableTree(node.tag,
int(node.attrib.get('colspan', '1')),
int(node.attrib.get('rowspan', '1')),
cell, *deque())
else:
new_node = TableTree(node.tag, None, None, None, *deque())
if parent is not None:
parent.children.append(new_node)
if node.tag != 'td':
for n in node.getchildren():
self.load_html_tree(n, new_node)
if parent is None:
return new_node
def evaluate(self, pred, true):
''' Computes TEDS score between the prediction and the ground truth of a
given sample
'''
if (not pred) or (not true):
return 0.0
parser = html.HTMLParser(remove_comments=True, encoding='utf-8')
pred = html.fromstring(pred, parser=parser)
true = html.fromstring(true, parser=parser)
if pred.xpath('body/table') and true.xpath('body/table'):
pred = pred.xpath('body/table')[0]
true = true.xpath('body/table')[0]
if self.ignore_nodes:
etree.strip_tags(pred, *self.ignore_nodes)
etree.strip_tags(true, *self.ignore_nodes)
n_nodes_pred = len(pred.xpath(".//*"))
n_nodes_true = len(true.xpath(".//*"))
n_nodes = max(n_nodes_pred, n_nodes_true)
tree_pred = self.load_html_tree(pred)
tree_true = self.load_html_tree(true)
distance = APTED(tree_pred, tree_true,
CustomConfig()).compute_edit_distance()
return 1.0 - (float(distance) / n_nodes)
else:
return 0.0
def batch_evaluate(self, pred_json, true_json):
''' Computes TEDS score between the prediction and the ground truth of
a batch of samples
@params pred_json: {'FILENAME': 'HTML CODE', ...}
@params true_json: {'FILENAME': {'html': 'HTML CODE'}, ...}
@output: {'FILENAME': 'TEDS SCORE', ...}
'''
samples = true_json.keys()
if self.n_jobs == 1:
scores = [self.evaluate(pred_json.get(
filename, ''), true_json[filename]['html']) for filename in tqdm(samples)]
else:
inputs = [{'pred': pred_json.get(
filename, ''), 'true': true_json[filename]['html']} for filename in samples]
scores = parallel_process(
inputs, self.evaluate, use_kwargs=True, n_jobs=self.n_jobs, front_num=1)
scores = dict(zip(samples, scores))
return scores
def batch_evaluate_html(self, pred_htmls, true_htmls):
''' Computes TEDS score between the prediction and the ground truth of
a batch of samples
'''
if self.n_jobs == 1:
scores = [self.evaluate(pred_html, true_html) for (
pred_html, true_html) in zip(pred_htmls, true_htmls)]
else:
inputs = [{"pred": pred_html, "true": true_html} for(
pred_html, true_html) in zip(pred_htmls, true_htmls)]
scores = parallel_process(
inputs, self.evaluate, use_kwargs=True, n_jobs=self.n_jobs, front_num=1)
return scores
if __name__ == '__main__':
import json
import pprint
with open('sample_pred.json') as fp:
pred_json = json.load(fp)
with open('sample_gt.json') as fp:
true_json = json.load(fp)
teds = TEDS(n_jobs=4)
scores = teds.batch_evaluate(pred_json, true_json)
pp = pprint.PrettyPrinter()
pp.pprint(scores)
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
\ No newline at end of file
此差异已折叠。
此差异已折叠。
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from PIL import Image
import numpy as np
from tools.infer.utility import draw_ocr_box_txt, init_args as infer_args
def init_args():
parser = infer_args()
# params for output
parser.add_argument("--output", type=str, default='./output/table')
# params for table structure
parser.add_argument("--structure_max_len", type=int, default=488)
parser.add_argument("--structure_max_text_length", type=int, default=100)
parser.add_argument("--structure_max_elem_length", type=int, default=800)
parser.add_argument("--structure_max_cell_num", type=int, default=500)
parser.add_argument("--structure_model_dir", type=str)
parser.add_argument("--structure_char_type", type=str, default='en')
parser.add_argument("--structure_char_dict_path", type=str, default="../ppocr/utils/dict/table_structure_dict.txt")
# params for layout detector
parser.add_argument("--layout_model_dir", type=str)
return parser
def parse_args():
parser = init_args()
return parser.parse_args()
def draw_result(image, result, font_path):
if isinstance(image, np.ndarray):
image = Image.fromarray(image)
boxes, txts, scores = [], [], []
for region in result:
if region['type'] == 'Table':
pass
elif region['type'] == 'Figure':
pass
else:
for box, rec_res in zip(region['res'][0], region['res'][1]):
boxes.append(np.array(box).reshape(-1, 2))
txts.append(rec_res[0])
scores.append(rec_res[1])
im_show = draw_ocr_box_txt(image, boxes, txts, scores, font_path=font_path,drop_score=0)
return im_show
\ No newline at end of file
......@@ -43,7 +43,7 @@ class TextDetector(object):
pre_process_list = [{
'DetResizeForTest': {
'limit_side_len': args.det_limit_side_len,
'limit_type': args.det_limit_type
'limit_type': args.det_limit_type,
}
}, {
'NormalizeImage': {
......
此差异已折叠。
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册