layout_parsing_v2.yaml 3.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131
  1. pipeline_name: layout_parsing_v2
  2. use_doc_preprocessor: True
  3. use_general_ocr: True
  4. use_seal_recognition: True
  5. use_table_recognition: True
  6. use_formula_recognition: True
  7. SubModules:
  8. LayoutDetection:
  9. module_name: layout_detection
  10. model_name: PP-DocLayout-L
  11. model_dir: null
  12. threshold:
  13. 7: 0.3
  14. layout_nms: True
  15. layout_unclip_ratio: 1.0
  16. layout_merge_bboxes_mode: "large"
  17. SubPipelines:
  18. DocPreprocessor:
  19. pipeline_name: doc_preprocessor
  20. use_doc_orientation_classify: True
  21. use_doc_unwarping: True
  22. SubModules:
  23. DocOrientationClassify:
  24. module_name: doc_text_orientation
  25. model_name: PP-LCNet_x1_0_doc_ori
  26. model_dir: null
  27. DocUnwarping:
  28. module_name: image_unwarping
  29. model_name: UVDoc
  30. model_dir: null
  31. GeneralOCR:
  32. pipeline_name: OCR
  33. text_type: general
  34. use_doc_preprocessor: False
  35. use_textline_orientation: True
  36. SubModules:
  37. TextDetection:
  38. module_name: text_detection
  39. model_name: PP-OCRv4_mobile_det
  40. model_dir: null
  41. limit_side_len: 960
  42. limit_type: max
  43. thresh: 0.3
  44. box_thresh: 0.6
  45. unclip_ratio: 2.0
  46. TextLineOrientation:
  47. module_name: textline_orientation
  48. model_name: PP-LCNet_x0_25_textline_ori
  49. model_dir: null
  50. batch_size: 1
  51. TextRecognition:
  52. module_name: text_recognition
  53. model_name: PP-OCRv4_server_rec_doc
  54. model_dir: null
  55. batch_size: 1
  56. score_thresh: 0.0
  57. TableRecognition:
  58. pipeline_name: table_recognition_v2
  59. use_layout_detection: False
  60. use_doc_preprocessor: False
  61. use_ocr_model: False
  62. SubModules:
  63. TableClassification:
  64. module_name: table_classification
  65. model_name: PP-LCNet_x1_0_table_cls
  66. model_dir: null
  67. WiredTableStructureRecognition:
  68. module_name: table_structure_recognition
  69. model_name: SLANeXt_wired
  70. model_dir: null
  71. WirelessTableStructureRecognition:
  72. module_name: table_structure_recognition
  73. model_name: SLANeXt_wireless
  74. model_dir: null
  75. WiredTableCellsDetection:
  76. module_name: table_cells_detection
  77. model_name: RT-DETR-L_wired_table_cell_det
  78. model_dir: null
  79. WirelessTableCellsDetection:
  80. module_name: table_cells_detection
  81. model_name: RT-DETR-L_wireless_table_cell_det
  82. model_dir: null
  83. SealRecognition:
  84. pipeline_name: seal_recognition
  85. use_layout_detection: False
  86. use_doc_preprocessor: False
  87. SubPipelines:
  88. SealOCR:
  89. pipeline_name: OCR
  90. text_type: seal
  91. use_doc_preprocessor: False
  92. use_textline_orientation: False
  93. SubModules:
  94. TextDetection:
  95. module_name: seal_text_detection
  96. model_name: PP-OCRv4_server_seal_det
  97. model_dir: null
  98. limit_side_len: 736
  99. limit_type: min
  100. thresh: 0.2
  101. box_thresh: 0.6
  102. unclip_ratio: 0.5
  103. TextRecognition:
  104. module_name: text_recognition
  105. model_name: PP-OCRv4_server_rec
  106. model_dir: null
  107. batch_size: 1
  108. score_thresh: 0
  109. FormulaRecognition:
  110. pipeline_name: formula_recognition
  111. use_layout_detection: False
  112. use_doc_preprocessor: False
  113. SubModules:
  114. FormulaRecognition:
  115. module_name: formula_recognition
  116. model_name: PP-FormulaNet-L
  117. model_dir: null
  118. batch_size: 5