优化ocr模型，增加关键词

photo_mask.py

@@ -9,7 +9,7 @@ from time import sleep
 import cv2
 import numpy as np
 import paddleclas
-from paddleocr import PaddleOCR
+from paddlenlp.utils.doc_parser import DocParser
 from sqlalchemy import update
 
 from auto_email.error_email import send_an_error_email
@@ -20,7 +20,7 @@ from config.ucloud import BUCKET
 from models import ZxPhrec, ZxPhhd
 from ucloud import ucloud
 
-OCR = PaddleOCR(use_angle_cls=False, lang="ch", show_log=False, gpu_id=1)
+DOC_PARSER = DocParser(use_gpu=True, device_id=1)
 
 
 def open_image(img_path):
@@ -258,7 +258,8 @@ def get_mask_layout(image, content):
 
     result = []
     try:
-        layouts = get_ocr_layout(OCR, temp_file.name)
+        # layouts = get_ocr_layout(OCR, temp_file.name)
+        layouts = DOC_PARSER.parse({"doc": temp_file.name})["layout"]
         if not layouts:
             # 无识别结果
             return result
@@ -273,6 +274,8 @@ def get_mask_layout(image, content):
                     result.append(find_box_of_value("交款人", layout))
                 if "文款人" in layout[1]:
                     result.append(find_box_of_value("文款人", layout))
+                if "购买方名称" in layout[1]:
+                    result.append(find_box_of_value("购买方名称", layout))
             return result
     except Exception as e:
         logging.error("涂抹时出错", exc_info=e)