Browse Source

Merge pull request #1088 from myhloli/dev

feat(ocr): filter out low confidence ocr results
Xiaomeng Zhao 1 year ago
parent
commit
9675a57416
1 changed files with 3 additions and 0 deletions
  1. 3 0
      magic_pdf/model/sub_modules/ocr/paddleocr/ocr_utils.py

+ 3 - 0
magic_pdf/model/sub_modules/ocr/paddleocr/ocr_utils.py

@@ -214,6 +214,9 @@ def get_ocr_result_list(ocr_res, useful_list):
         if len(box_ocr_res) == 2:
         if len(box_ocr_res) == 2:
             p1, p2, p3, p4 = box_ocr_res[0]
             p1, p2, p3, p4 = box_ocr_res[0]
             text, score = box_ocr_res[1]
             text, score = box_ocr_res[1]
+            # logger.info(f"text: {text}, score: {score}")
+            if score < 0.6:  # 过滤低置信度的结果
+                continue
         else:
         else:
             p1, p2, p3, p4 = box_ocr_res
             p1, p2, p3, p4 = box_ocr_res
             text, score = "", 1
             text, score = "", 1