Library Structure

Overview of the manuscript-ocr library architecture.

        graph LR
    %% Main package
    manuscript[manuscript]

    %% Core modules
    manuscript --> Pipeline["Pipeline<br/><i>Главный класс для OCR</i>"]
    manuscript --> data["data<br/><i>Структуры данных</i>"]
    manuscript --> detectors["detectors<br/><i>Детекторы текста</i>"]
    manuscript --> recognizers["recognizers<br/><i>Распознаватели текста</i>"]
    manuscript --> utils["utils<br/><i>Утилиты</i>"]
    manuscript --> api["api<br/><i>API базовые классы</i>"]

    %% Pipeline methods
    Pipeline --> p_predict["predict()<br/><i>→ Dict | Tuple[Dict, Image]</i>"]
    Pipeline --> p_get_text["get_text()<br/><i>→ str</i>"]

    %% Data structures
    data --> Page["Page"]
    data --> Block["Block"]
    data --> Line["Line"]
    data --> Word["Word"]

    Page --> page_blocks["blocks: List[Block]"]
    Block --> block_lines["lines: List[Line]"]
    Block --> block_order["order: Optional[int]"]
    Line --> line_words["words: List[Word]"]
    Line --> line_order["order: Optional[int]"]
    Word --> word_polygon["polygon: List[(x,y)]"]
    Word --> word_det_conf["detection_confidence: float"]
    Word --> word_text["text: Optional[str]"]
    Word --> word_rec_conf["recognition_confidence: Optional[float]"]
    Word --> word_order["order: Optional[int]"]

    %% Detectors
    detectors --> EAST["EAST<br/><i>Efficient and Accurate Scene Text Detector</i>"]
    EAST --> east_predict["predict()<br/><i>→ Dict[str, Any]</i>"]
    EAST --> east_train["train()<br/><i>→ None</i>"]
    EAST --> east_export["export()<br/><i>→ str</i>"]

    %% Recognizers
    recognizers --> TRBA["TRBA<br/><i>Text Recognition with BiLSTM + Attention</i>"]
    TRBA --> trba_predict["predict()<br/><i>→ List[Dict[str, Any]]</i>"]
    TRBA --> trba_train["train()<br/><i>→ None</i>"]
    TRBA --> trba_export["export()<br/><i>→ str</i>"]

    %% Utils submodules
    utils --> io["io<br/><i>Чтение/запись</i>"]
    utils --> visualization["visualization<br/><i>Визуализация результатов</i>"]
    utils --> sorting["sorting<br/><i>Сортировка и организация</i>"]
    utils --> training["training<br/><i>Обучение моделей</i>"]

    %% IO functions
    io --> read_image["read_image()<br/><i>→ np.ndarray</i>"]

    %% Visualization functions
    visualization --> visualize_page["visualize_page()<br/><i>→ Image</i>"]

    %% Sorting functions
    sorting --> organize_page["organize_page()<br/><i>→ Page</i>"]

    %% Training functions
    training --> set_seed["set_seed()<br/><i>→ None</i>"]

    %% API
    api --> BaseModel["BaseModel<br/><i>Базовый класс для моделей</i>"]
    BaseModel --> base_predict["predict() <i>abstract</i>"]

    %% Styles
    style manuscript fill:#1f2937,color:#ffffff,stroke:#111827,stroke-width:2px
    style Pipeline fill:#fde68a,color:#111827,stroke:#92400e,stroke-width:2px

    style data fill:#bbf7d0,color:#064e3b,stroke:#047857
    style detectors fill:#bfdbfe,color:#1e3a8a,stroke:#2563eb
    style recognizers fill:#ddd6fe,color:#4c1d95,stroke:#7c3aed
    style utils fill:#fed7aa,color:#7c2d12,stroke:#ea580c
    style api fill:#fecaca,color:#7f1d1d,stroke:#dc2626

Module Descriptions

Pipeline

The main high-level interface that combines detection and recognition into a single OCR workflow.

data

Data structures (Page, Block, Line, Word) for representing OCR results in a hierarchical format.

detectors

Text detection models. Currently includes EAST (Efficient and Accurate Scene Text Detector).

recognizers

Text recognition models. Currently includes TRBA (Text Recognition with BiLSTM and Attention mechanism).

utils

Utility functions for:

I/O operations (read_image)
Visualization (visualize_page)
Sorting and organization (organize_page)
Training utilities (set_seed)