Library Structure

        graph LR
    manuscript[manuscript] --> Pipeline
    manuscript --> Stack
    manuscript --> data
    manuscript --> detectors
    manuscript --> layouts
    manuscript --> recognizers
    manuscript --> correctors
    manuscript --> models
    manuscript --> utils
    manuscript --> api

    data --> elements["Page, Element, TextRegion<br/>Table, Cell, Image"]
    detectors --> detection["YOLO, EAST, PPOCR<br/>RFDETR, Mask2Former"]
    layouts --> ordering["SimpleSorting<br/>ReadingSorting"]
    recognizers --> recognition["TRBA, PPOCRRec, TrOCR"]
    recognizers --> ensembles["MajorityVotingRecognizer<br/>MedoidRecognizer"]
    correctors --> CharLM
    models --> registry["Registry and artifact cache"]
    api --> contracts["Base classes and protocols"]
    

Modules

Module or class

Responsibility

Pipeline

Runs configured stages and exposes Page snapshots and execution timings.

Stack

Composes processors, accumulates detector regions and optionally runs independent detectors in parallel.

data

Defines elements, parent-child links, validation, COCO serialization and text output.

detectors

Produces configured regions and optional detailed detection results.

layouts

Groups, reparents and orders existing regions.

recognizers

Recognizes selected text regions and combines independent predictions.

correctors

Updates recognized text, including CharLM character substitutions.

models

Resolves remote catalogs and manages local model artifacts.

utils

Provides image input, geometry, grouping, transformations and visualization.

api

Defines stage protocols, base classes and shared artifact-loading behavior.

Public entry points

Import the components you need from their packages:

from manuscript import Pipeline, Stack
from manuscript.data import Page, TextRegion
from manuscript.detectors import YOLO
from manuscript.layouts import SimpleSorting, ReadingSorting
from manuscript.recognizers import TRBA, MajorityVotingRecognizer
from manuscript.correctors import CharLM
from manuscript import models