layout_parsing_v2.yaml 3.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137
  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: False
  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. TextRecognition:
  47. module_name: text_recognition
  48. model_name: PP-OCRv4_server_rec
  49. model_dir: null
  50. batch_size: 1
  51. score_thresh: 0.0
  52. TableRecognition:
  53. pipeline_name: table_recognition_v2
  54. use_layout_detection: False
  55. use_doc_preprocessor: False
  56. use_ocr_model: False
  57. SubModules:
  58. TableClassification:
  59. module_name: table_classification
  60. model_name: PP-LCNet_x1_0_table_cls
  61. model_dir: null
  62. WiredTableStructureRecognition:
  63. module_name: table_structure_recognition
  64. model_name: SLANeXt_wired
  65. model_dir: null
  66. WirelessTableStructureRecognition:
  67. module_name: table_structure_recognition
  68. model_name: SLANeXt_wireless
  69. model_dir: null
  70. WiredTableCellsDetection:
  71. module_name: table_cells_detection
  72. model_name: RT-DETR-L_wired_table_cell_det
  73. model_dir: null
  74. WirelessTableCellsDetection:
  75. module_name: table_cells_detection
  76. model_name: RT-DETR-L_wireless_table_cell_det
  77. model_dir: null
  78. # TableRecognition:
  79. # pipeline_name: table_recognition
  80. # use_layout_detection: False
  81. # use_doc_preprocessor: False
  82. # use_ocr_model: False
  83. # SubModules:
  84. # TableStructureRecognition:
  85. # module_name: table_structure_recognition
  86. # model_name: SLANet_plus
  87. # model_dir: null
  88. SealRecognition:
  89. pipeline_name: seal_recognition
  90. use_layout_detection: False
  91. use_doc_preprocessor: False
  92. SubPipelines:
  93. SealOCR:
  94. pipeline_name: OCR
  95. text_type: seal
  96. use_doc_preprocessor: False
  97. use_textline_orientation: False
  98. SubModules:
  99. TextDetection:
  100. module_name: seal_text_detection
  101. model_name: PP-OCRv4_server_seal_det
  102. model_dir: null
  103. limit_side_len: 736
  104. limit_type: min
  105. thresh: 0.2
  106. box_thresh: 0.6
  107. unclip_ratio: 0.5
  108. TextRecognition:
  109. module_name: text_recognition
  110. model_name: PP-OCRv4_server_rec
  111. model_dir: null
  112. batch_size: 1
  113. score_thresh: 0
  114. FormulaRecognition:
  115. pipeline_name: formula_recognition
  116. use_layout_detection: False
  117. use_doc_preprocessor: False
  118. SubModules:
  119. FormulaRecognition:
  120. module_name: formula_recognition
  121. model_name: PP-FormulaNet-L
  122. model_dir: null
  123. batch_size: 5