Echo/fcb_photo_review
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

调整涂抹的ocr模型参数；优化关键词

Browse Source
This commit is contained in:
liuyebo
2024-07-19 12:15:02 +08:00
parent 69f4fecd4f
commit 5193839c9e
2 changed files with 13 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
photo_mask/__init__.py
View File

@@ -1,4 +1,4 @@
from paddlenlp.utils.doc_parser import DocParser
from paddleocr import PaddleOCR
"""
项目配置
@@ -19,11 +19,11 @@ UPLOAD_TRY_TIMES = 3
"""
NAME_KEYS = [
{"key": "姓名", "length": 4},
{"key": "交款人", "length": 4},
{"key": "文款人", "length": 4},
{"key": "娃名", "length": 4},
{"key": "款人", "length": 4},
{"key": "购买方名称", "length": 4},
]
ID_CARD_NUM_KEYS = [{"key": "身份证号", "length": 19}, ]
# OCR = PaddleOCR(use_angle_cls=False, show_log=False, gpu_id=1)
OCR = DocParser(use_gpu=True, layout_analysis=False)
# 如果不希望识别出空格，可以设置use_space_char=False。做此项设置一定要测试，2.7.3版本此项设置有bug，会导致识别失败
OCR = PaddleOCR(use_angle_cls=False, show_log=False, det_db_box_thresh=0.3)

13
photo_mask/photo_mask.py
View File

@@ -60,7 +60,7 @@ def find_boxes(content, layout, offset=0, length=None, improve=False, image_path
captured_image, offset_x, offset_y = image_util.expand_to_a4_size(captured_image, True)
cv2.imwrite(temp_file.name, captured_image)
try:
layouts = OCR.parse({"doc": temp_file.name})["layout"]
layouts = util.get_ocr_layout(OCR, temp_file.name)
except TypeError:
# 如果是类型错误，大概率是没识别到文字
layouts = []
@@ -91,8 +91,8 @@ def get_mask_layout(image, name, id_card_num):
result = []
try:
try:
# layouts = util.get_ocr_layout(OCR, temp_file.name)
layouts = OCR.parse({"doc": temp_file.name})["layout"]
layouts = util.get_ocr_layout(OCR, temp_file.name)
# layouts = OCR.parse({"doc": temp_file.name})["layout"]
except TypeError:
# 如果是类型错误，大概率是没识别到文字
layouts = []
@@ -118,6 +118,10 @@ def get_mask_layout(image, name, id_card_num):
keys = []
if find_name_by_key:
keys += NAME_KEYS
if layout[1].startswith("名："):
result += find_boxes("名：", layout, offset=2, length=3)
if layout[1].startswith("名:"):
result += find_boxes("名:", layout, offset=2, length=3)
if find_id_card_num_by_key:
keys += ID_CARD_NUM_KEYS
for key in keys:
@@ -135,8 +139,7 @@ def handle_image_for_mask(split_result):
expand_img, offset_x, offset_y = image_util.expand_to_a4_size(split_result["img"], True)
split_result["x_offset"] -= offset_x
split_result["y_offset"] -= offset_y
gray_image = cv2.cvtColor(expand_img, cv2.COLOR_BGR2GRAY)
return gray_image, split_result["x_offset"], split_result["y_offset"]
return expand_img, split_result["x_offset"], split_result["y_offset"]
def photo_mask(pk_phhd, name, id_card_num):