points 是框选的矩形四个角坐标： 左上 右上 右下 左下

main.py

@@ -180,6 +180,17 @@ def ocr_one_pdf(pth_pdf):
     reader = PdfReader(pth_pdf)
     number_of_pages = len(reader.pages)
 
+    # see huggingface/PPOCRLabel/PPOCRLabel.py   for image notation
+    name_pp_label = 'Label.txt'
+    pth_pp_label = os.path.join(dir, name_pp_label)
+    pp_label_text = ''
+        # real/0010.jpg	[{"transcription": "待识别", "points": [[137, 77], [740, 77], [740, 165], [137, 165]], "difficult": false}]
+        # points 是框选的矩形四个角坐标： 左上 右上  右下 左下
+    name_pp_state = 'fileState.txt'
+    pth_pp_state = os.path.join(dir, name_pp_state)
+    pp_state_text = ''
+        # D:\workcode\nodejs\框选问题\real\0010.jpg	1
+    
 
     for nth_page in range(1, number_of_pages+1):
 
@@ -270,6 +281,7 @@ def ocr_one_pdf(pth_pdf):
         print( f'one page done. {nth_page} / {number_of_pages}' )
 
 
+
 if __name__ == '__main__':
 
     ocr_one_pdf('pdfs/jp/高木直子学日语文法不要想太多/高木直子学日语文法不要想太多.pdf')