rednote-hilab/dots.ocr
Dots.ocr
dots.ocr is a suite of software utilities for document layout analysis, multilingual optical character recognition, and scene text digitization. It functions as an engine for extracting digital text and structured layout data from images and PDFs across various human scripts.
The project includes a specialized transformer for converting charts, diagrams, and chemical formulas from raster images into scalable vector graphics. It also provides a pipeline to transform extracted text and structural layout from documents and web screenshots into formatted Markdown files.
The system covers capabilities for identifying bounding boxes and categories of layout elements to produce structured JSON representations. It further includes tools for scene text detection within natural images and an evaluation framework for measuring text and table extraction accuracy against ground truth data.
Features
- Document Analysis Tools - Provides a comprehensive tool for detecting bounding boxes and layout categories to structure documents as JSON.
- Bounding Box Detection - Provides bounding box detection to locate page elements and distinguish body text from headers and footers.
- Multilingual OCR Systems - Extracts digital text and structured layout data from images and PDFs across various human scripts.
- Document Layout Analysis - Identifies bounding boxes and categories for layout elements within images and PDF files.
- Multilingual Text Recognition - Implements a unified recognition engine to process diverse character sets and writing systems from images into digital text.
- Scene Text Recognition - Locates and extracts digital text from natural images and complex environmental backgrounds.
- OCR-Based Extraction - Extracts digital text and structured layout data from images and PDFs across various human scripts.
- Scene Text Detection Systems - Finds and digitizes text embedded within natural scene images and web screenshots.
- Document Digitization Tools - Converts images and PDFs containing various human scripts into digital text while preserving document structure.
- Automated Generation - Transforms extracted text and structural layout from documents and web screenshots into formatted Markdown files.
- Document Layout Analyzers - Identifies bounding boxes and categories of layout elements to produce structured JSON representations of documents.
- Markdown Converters - Transforms extracted text and structural layout from images and documents into formatted Markdown files.
- Visual-to-Markdown Pipelines - Transforms scanned documents and web screenshots into formatted Markdown files for editing and archival.
- Layout-Aware Extraction - Maps spatial relationships between text elements and document structures during the extraction process.
- Layout-Preserving Export - Generates formatted Markdown files from extracted document text while excluding page headers and footers.
- Raster-to-Vector Conversions - Transforms rasterized charts, diagrams, and chemical formulas into scalable vector graphics.
- Model Serving Engines - Implements high-performance model serving engines to provide scalable text extraction API endpoints.
- Model Serving Servers - Serves machine learning models as scalable API endpoints using high-performance inference servers.
- Text Detection Models - Locates and digitizes text embedded within natural scene images and complex backgrounds.
- OCR Accuracy Evaluators - Measures text and table extraction accuracy against ground truth images to refine OCR performance.
- Visual Layout Parsing - Extracts structured text and layout information specifically from screenshots of web pages.
- Vector Graphics Export - Converts technical graphics like charts and chemical formulas from images into scalable vector graphics.
- Raster-to-SVG Transformers - Converts charts, diagrams, and chemical formulas from raster images into scalable vector graphics code.
