draw_bbox.py 2.8 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071
  1. from magic_pdf.libs.commons import fitz # PyMuPDF
  2. def draw_bbox_without_number(i, bbox_list, page, rgb_config):
  3. new_rgb = []
  4. for item in rgb_config:
  5. item = float(item) / 255
  6. new_rgb.append(item)
  7. page_data = bbox_list[i]
  8. for bbox in page_data:
  9. x0, y0, x1, y1 = bbox
  10. rect_coords = fitz.Rect(x0, y0, x1, y1) # Define the rectangle
  11. page.draw_rect(rect_coords, color=new_rgb, fill=None, width=0.5, overlay=True) # Draw the rectangle
  12. def draw_bbox_with_number(i, bbox_list, page, rgb_config):
  13. new_rgb = []
  14. for item in rgb_config:
  15. item = float(item) / 255
  16. new_rgb.append(item)
  17. page_data = bbox_list[i]
  18. for j, bbox in enumerate(page_data):
  19. x0, y0, x1, y1 = bbox
  20. rect_coords = fitz.Rect(x0, y0, x1, y1) # Define the rectangle
  21. page.draw_rect(rect_coords, color=new_rgb, fill=None, width=0.5, overlay=True) # Draw the rectangle
  22. page.insert_text((x0, y0), str(j + 1), fontsize=10, color=new_rgb) # Insert the index at the top left corner of the rectangle
  23. def draw_layout_bbox(pdf_info_dict, input_path, out_path):
  24. layout_bbox_list = []
  25. for page in pdf_info_dict.values():
  26. page_list = []
  27. for layout in page['layout_bboxes']:
  28. page_list.append(layout['layout_bbox'])
  29. layout_bbox_list.append(page_list)
  30. doc = fitz.open(input_path)
  31. for i, page in enumerate(doc):
  32. draw_bbox_with_number(i, layout_bbox_list, page, [255, 0, 0])
  33. # Save the PDF
  34. doc.save(f"{out_path}/layout.pdf")
  35. def draw_text_bbox(pdf_info_dict, input_path, out_path):
  36. text_list = []
  37. inline_equation_list = []
  38. displayed_equation_list = []
  39. for page in pdf_info_dict.values():
  40. page_text_list = []
  41. page_inline_equation_list = []
  42. page_displayed_equation_list = []
  43. for block in page['preproc_blocks']:
  44. for line in block['lines']:
  45. for span in line['spans']:
  46. if span['type'] == 'text':
  47. page_text_list.append(span['bbox'])
  48. elif span['type'] == 'inline_equation':
  49. page_inline_equation_list.append(span['bbox'])
  50. elif span['type'] == 'displayed_equation':
  51. page_displayed_equation_list.append(span['bbox'])
  52. text_list.append(page_text_list)
  53. inline_equation_list.append(page_inline_equation_list)
  54. displayed_equation_list.append(page_displayed_equation_list)
  55. doc = fitz.open(input_path)
  56. for i, page in enumerate(doc):
  57. # 获取当前页面的数据
  58. draw_bbox_without_number(i, text_list, page, [255, 0, 0])
  59. draw_bbox_without_number(i, inline_equation_list, page, [0, 255, 0])
  60. draw_bbox_without_number(i, displayed_equation_list, page, [0, 0, 255])
  61. # Save the PDF
  62. doc.save(f"{out_path}/text.pdf")