zhengchun

zhengchun извршује push на main у zhengchun/ocr_platform

  • 6913675caf fix(pipeline_manager): ensure bbox coordinates are integers for text PDFs - Updated the bounding box (bbox) assignment in the EnhancedDocPipeline to maintain integer types for coordinates, aligning with the integer pixel coordinates of PDF formats. This change enhances the accuracy of spatial data representation in document processing.
  • e8c4c45e38 feat(config): add new data source configuration for 山西云集科技有限公司 - Introduced a new YAML configuration file for 山西云集科技有限公司, detailing the data source settings for bank statement processing. - Included parameters for JSON and image directory patterns, output structure, and batch processing options to enhance data handling capabilities.
  • Поређење ових 2 комита »

пре 12 часа

zhengchun извршује push на main у zhengchun/ocr_platform

  • da44e4df80 feat(docs): update usage scenarios in README for OCR tools and document processing - Revised the usage scenarios for bank statement recognition and financial report processing to include the `universal_doc_parser`, `ocr_merger`, and `ocr_validator`. - Clarified the sequence of operations and enhanced descriptions for better understanding of the document processing workflow.
  • c86d45aabd feat(docs): add comprehensive documentation for Docling, PaddleX, and MinerU layout models - Introduced a new markdown file detailing the architecture, technical features, and comparative analysis of three layout models: Docling (RT-DETR), PaddleX (PicoDet/RT-DETR), and MinerU (DocLayout-YOLO). - Included sections on model architecture, performance metrics, and application scenarios to aid users in selecting the appropriate model for their needs.
  • 05a4aea5aa fix(docs): update title for clarity in Docling RT-DETR vs DiT comparison
  • Поређење ових 3 комита »

пре 16 часа

zhengchun извршује push на main у zhengchun/ocr_platform

  • d35da8581f feat(config): add new data source for OCR validation - Included a new data source entry for "提取自赤峰黄金2023年报.yaml" in the global configuration file to enhance OCR validation capabilities.
  • a3174f90d4 feat(grid_recovery): enhance line processing with geometric calculations and filtering methods - Introduced several static methods for geometric calculations, including line fitting, point-to-line distance, and distance between points. - Added functionality to extend lines for intersection and filter lines based on proximity to image edges and length thresholds. - Implemented methods to filter lines based on alignment with bounding boxes, improving the accuracy of line detection in document processing. - Updated debug image saving functionality to streamline debugging processes during line adjustments.
  • f4ec65b6b3 feat(config): update OCR tool configurations for bank statement processing - Modified the image directory path for the existing "mineru" tool in the bank_statement_yusys_v2 configuration. - Added a new configuration file for the "mineru" tool specific to bank_statement_wired_unet, including its result and image directories, along with a description and enablement option.
  • b38b4def56 feat(text_filling): enhance overlap detection logic for cell merging - Added calculation for cell overlap ratio to improve accuracy in determining overlapping cells. - Adjusted overlap conditions to implement dual thresholds for horizontal and vertical overlaps, enhancing robustness in text filling scenarios.
  • Поређење ових 4 комита »

пре 1 дан

zhengchun извршује push на main у zhengchun/ocr_platform

  • b18ddfa46e feat(config): add new OCR tool configuration for bank statement processing - Introduced a new OCR tool configuration for "mineru" with a dedicated result directory and image directory for bank_statement_wired_unet. - Enhanced documentation to include descriptions and enablement options for the new tool.
  • 83c6b132d5 feat(universal_doc_parser): enhance document processing with modular design and automatic table classification - Introduced a modular framework allowing flexible replacement of processing components for various scenarios. - Unified output format across different model combinations to ensure consistency in downstream processing. - Added automatic table classification feature to distinguish between wired and wireless tables, improving recognition accuracy. - Enhanced the main processing flow with detailed handling for both scanned documents and text PDFs, including fallback mechanisms for OCR. - Updated documentation to reflect new features and configuration options.
  • Поређење ових 2 комита »

пре 1 дан

zhengchun извршује push на main у zhengchun/ocr_platform

  • 4afb8f9fa5 feat(main_v2): 添加环境变量打印和参数验证功能,优化处理流水线创建逻辑

пре 2 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • 712d004e85 refactor(pdf_coordinate_transform): remove outdated bbox transformation function - Deleted the old `transform_bbox_for_rotation_pypdfium2_old` function to streamline the codebase. - This function was redundant and not utilized in the current implementation.
  • d560cc01e0 fix(paddle_vl_adapter): update model name retrieval logic to ensure default value is used - Changed model name assignment to use config.get() for better flexibility. - Ensured that the default model name 'PaddleOCR-VL-0.9B' is applied if not explicitly set in the configuration.
  • 11dcec8769 feat(paddle_table_classifier): integrate debugging tools for enhanced table line detection visualization - Added WiredTableVisualizer and WiredTableDebugUtils for improved debugging capabilities. - Updated classify and _detect_table_lines methods to incorporate debug options for better visualization control. - Merged debug options to allow for dynamic adjustments during classification.
  • ac8cca5d22 feat(pipeline): add debug options for table classification to enhance visualization control
  • 162d4eb7c2 fix(config): explicit set VLM model name in yusys_v2 config - Add `model_name: "PaddleOCR-VL-0.9B"` to bank_statement_yusys_v2.yaml to ensure correct model usage.
  • Поређење ових 7 комита »

пре 2 дана

zhengchun извршује push на main у zhengchun/ocr_platform

пре 2 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • 985c776438 feat: 添加自定义Numpy编码器以支持JSON格式化,优化输出中的numpy数据处理
  • 69f11bfbcd feat: 增强网格结构恢复逻辑,支持OCR边缘补偿功能,传递OCR结果以优化单元格提取
  • 87212bfd61 feat: 增加OCR边缘补偿功能,支持未封闭单元格的补偿,优化网格结构恢复逻辑
  • 7a99597a84 feat: 增强文档处理功能,支持OCR辅助合并单元格,优化输出配置逻辑
  • f6c8b03d9c feat: 增加OCR补偿功能以恢复缺失单元格,优化网格结构重建逻辑
  • Поређење ових 5 комита »

пре 3 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • cb31f9e67f feat: 移除详细验证和调试日志,简化网格结构恢复逻辑

пре 4 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • f32729046f feat: 统一坐标系输出逻辑,增加return_upright_coords参数支持,优化文本提取功能
  • 44635af320 feat: 增强PDF旋转验证测试,添加return_upright_coords参数支持及结果对比功能
  • 10b6bdb6e4 feat: 更新页面方向识别逻辑,统一rotate_angle定义并处理文字PDF的旋转
  • 975ab2f230 feat: 更新二次OCR填充逻辑,增加OCR误合并检测和文本空值处理
  • ea2eab596b feat: 更新页面方向识别注释,明确rotate_angle用于图片旋转的角度
  • Поређење ових 8 комита »

пре 4 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • 0b7809226c feat: 添加PaddleOCR表格分类器适配器,支持有线/无线表格分类
  • 630cf15a2d feat: 添加表格分类器支持,优化表格识别路径选择
  • 57bcb4628d feat: 添加创建表格分类器的方法以区分有线/无线表格
  • 768858cbff feat: 添加可选的表格分类器参数以支持有线/无线表格的区分
  • a4b4be0968 feat: 添加布局后处理和表格分类配置,优化表格识别逻辑
  • Поређење ових 5 комита »

пре 5 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • c11f2ea045 feat: 添加 detect_page_type 方法以检测 PDF 页面类型(文本或扫描OCR)
  • f2079d9e4f feat: 添加 detect_page_type 函数以检测 PDF 页面的类型(文字页或图片页)
  • ca0374db5f feat: 添加 pdf_type 参数以支持不同的 PDF 处理模式,优化识别逻辑
  • 54f5b5943d fix: 将日志级别从 info 更改为 debug,以减少输出冗余
  • 9d7afeff31 feat: 添加 pdf_type 参数以支持不同的 OCR 模式,优化二次 OCR 逻辑
  • Поређење ових 8 комита »

пре 5 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • 58d9568b0f feat: add PDF text extraction module and refactor PDF utilities - Introduced `pdf_text_extraction.py` for extracting text from PDF documents using pypdfium2 and fitz engines. - Implemented functions for both region-specific and full-page text extraction, including automatic rotation handling. - Refactored `pdf_utils.py` to integrate new text extraction functionalities and maintain backward compatibility. - Organized PDF-related functionalities into submodules for better structure and maintainability. - Updated documentation to reflect changes and provide clear usage instructions.
  • e8a1f6ece8 feat: 添加 PDF 旋转验证测试程序,支持不同角度的文本坐标验证与渲染一致性
  • f86659f440 fix: 更新日志信息,包含倾斜阈值以便于调试
  • 752de6fa64 fix: 优化 OCR 重新识别逻辑,调整置信度阈值并移除不必要的旋转处理
  • 3b53b1f441 fix: 调整边缘过滤参数,修正 crop_padding 对应的计算逻辑
  • Поређење ових 7 комита »

пре 6 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • dd92babb27 feat: 改进文档处理流程,支持从 PDF 提取文本并与 OCR 结果对比,添加调试模式
  • dc9a615776 fix: 修复加载和分类文档时的参数传递,添加 renderer_used 参数
  • 85f5dfa1f4 feat: 更新 process_text_element 方法,改进预匹配 spans 的处理逻辑,支持 OCR 和 PDF 文本提取来源
  • 43d0e1c5d3 feat: 在 bank_statement_wired_unet 配置中启用调试模式
  • db1a81a141 feat: 添加 PDF 文档类型检测功能,支持 pypdfium2 和 fitz 渲染引擎,优化文本提取过程
  • Поређење ових 6 комита »

пре 6 дана

zhengchun извршује push на main у zhengchun/ocr_platform

  • ae0a19dc4d chore: Add .gitignore to exclude various development and output files, and update main_v2.py.
  • 481b5ea371 feat: add .gitignore to exclude common development artifacts, temporary files, and specific project outputs.
  • a4b8405df5 feat: 添加 bank_statement_wired_unet OCR 工具配置,支持结果目录和图像目录定义
  • Поређење ових 3 комита »

пре 1 недеља

zhengchun извршује push на main у zhengchun/ocr_platform

пре 1 недеља

zhengchun извршује push на main у zhengchun/ocr_platform

пре 1 недеља

zhengchun извршује push на main у zhengchun/ocr_platform

  • f75f1bb639 Fix typo in documentation regarding U-Net line detection and comparison with existing methods.
  • af1c467c48 fix: 优化UNet推理调试信息记录,增强尺寸一致性验证,返回检测到的倾斜角度
  • 6be6c8c3bb fix: 移除UNet预处理的宽高缩放因子参数,优化坐标转换的调试信息记录
  • ef447c6c7b fix: 从识别结果中获取倾斜角度,优化表格处理的准确性
  • Поређење ових 4 комита »

пре 1 недеља

zhengchun извршује push на main у zhengchun/ocr_platform

  • 64652051e4 fix: 更新示例输入输出路径,修正注释以提高代码可读性
  • ca720abd31 fix: 增强UNet预处理的缩放因子验证,优化预测结果的尺寸一致性检查,记录详细的调试信息以确保坐标转换的准确性
  • 1fbcf06f4a fix: 增强文本填充器的OCR检测能力,支持跨单元格检测和输出调试图像,优化重叠检测逻辑
  • bb0acb2afc fix: 增强网格结构恢复中的坐标转换精度,添加调试信息以验证缩放比例和单元格覆盖情况
  • 3cf3aa5085 fix: 调整表格处理中的padding策略,优化边缘保护与噪声控制的平衡
  • Поређење ових 7 комита »

пре 1 недеља

zhengchun извршује push на main у zhengchun/ocr_platform

  • 652b321bd6 feat: Update batch processing in main_v2.py to include output directory parameter for document processing, enhancing flexibility in file management.
  • 1bb438fba3 fix: Improve coordinate transformation accuracy in WiredTableVisualizer to reduce cumulative errors and enhance debugging with detailed logging of cell coordinates during visualization.
  • c628acd7b7 feat: Enhance text filling strategy in TextFiller class by introducing overlap ratio calculation for improved OCR box matching, optimizing performance with binary search, and refining text extraction logic.
  • 60aa86e4bf feat: Enhance GridRecovery class by adding optional parameters for original image dimensions and improving scaling logic for cell extraction, along with detailed debug logging for better traceability.
  • 26b500f344 feat: Add documentation for grid recovery scenarios in UNet, detailing causes of empty rows/columns and how `compress_grid` addresses them.
  • Поређење ових 6 комита »

пре 1 недеља