layout_parsing_v2.yaml 3.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132
  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. 0: 0.3
  14. 7: 0.3
  15. layout_nms: True
  16. layout_unclip_ratio: 1.0
  17. layout_merge_bboxes_mode: "large"
  18. SubPipelines:
  19. DocPreprocessor:
  20. pipeline_name: doc_preprocessor
  21. use_doc_orientation_classify: True
  22. use_doc_unwarping: True
  23. SubModules:
  24. DocOrientationClassify:
  25. module_name: doc_text_orientation
  26. model_name: PP-LCNet_x1_0_doc_ori
  27. model_dir: null
  28. DocUnwarping:
  29. module_name: image_unwarping
  30. model_name: UVDoc
  31. model_dir: null
  32. GeneralOCR:
  33. pipeline_name: OCR
  34. text_type: general
  35. use_doc_preprocessor: False
  36. use_textline_orientation: True
  37. SubModules:
  38. TextDetection:
  39. module_name: text_detection
  40. model_name: PP-OCRv4_mobile_det
  41. model_dir: null
  42. limit_side_len: 960
  43. limit_type: max
  44. thresh: 0.3
  45. box_thresh: 0.6
  46. unclip_ratio: 2.0
  47. TextLineOrientation:
  48. module_name: textline_orientation
  49. model_name: PP-LCNet_x0_25_textline_ori
  50. model_dir: null
  51. batch_size: 1
  52. TextRecognition:
  53. module_name: text_recognition
  54. model_name: PP-OCRv4_server_rec_doc
  55. model_dir: null
  56. batch_size: 6
  57. score_thresh: 0.0
  58. TableRecognition:
  59. pipeline_name: table_recognition_v2
  60. use_layout_detection: False
  61. use_doc_preprocessor: False
  62. use_ocr_model: False
  63. SubModules:
  64. TableClassification:
  65. module_name: table_classification
  66. model_name: PP-LCNet_x1_0_table_cls
  67. model_dir: null
  68. WiredTableStructureRecognition:
  69. module_name: table_structure_recognition
  70. model_name: SLANeXt_wired
  71. model_dir: null
  72. WirelessTableStructureRecognition:
  73. module_name: table_structure_recognition
  74. model_name: SLANeXt_wireless
  75. model_dir: null
  76. WiredTableCellsDetection:
  77. module_name: table_cells_detection
  78. model_name: RT-DETR-L_wired_table_cell_det
  79. model_dir: null
  80. WirelessTableCellsDetection:
  81. module_name: table_cells_detection
  82. model_name: RT-DETR-L_wireless_table_cell_det
  83. model_dir: null
  84. SealRecognition:
  85. pipeline_name: seal_recognition
  86. use_layout_detection: False
  87. use_doc_preprocessor: False
  88. SubPipelines:
  89. SealOCR:
  90. pipeline_name: OCR
  91. text_type: seal
  92. use_doc_preprocessor: False
  93. use_textline_orientation: False
  94. SubModules:
  95. TextDetection:
  96. module_name: seal_text_detection
  97. model_name: PP-OCRv4_server_seal_det
  98. model_dir: null
  99. limit_side_len: 736
  100. limit_type: min
  101. thresh: 0.2
  102. box_thresh: 0.6
  103. unclip_ratio: 0.5
  104. TextRecognition:
  105. module_name: text_recognition
  106. model_name: PP-OCRv4_server_rec
  107. model_dir: null
  108. batch_size: 1
  109. score_thresh: 0
  110. FormulaRecognition:
  111. pipeline_name: formula_recognition
  112. use_layout_detection: False
  113. use_doc_preprocessor: False
  114. SubModules:
  115. FormulaRecognition:
  116. module_name: formula_recognition
  117. model_name: PP-FormulaNet-L
  118. model_dir: null
  119. batch_size: 5