Digitize documents. Recognize handwritten content. Export structured text.

---

Type OCR Application

Input PDF / JPG / PNG

Output Text / JSON / PDF

Interface Streamlit

---

01 · Overview

Handwritten document digitization system ScribbleQ is an OCR-based document intelligence application for converting handwritten documents into clean digital text. The system combines document input handling, text-region detection, recognition, structured output, and an interactive Streamlit interface for fast document processing.

┌──────────────────────────────┐ │ SCRIBBLEQ OCR CONSOLE │ ├──────────────────────────────┤ │ Input PDF / Image │ │ Detect Text Regions │ │ Recognize Handwriting │ │ Output Text / JSON │ │ UI Streamlit │ └──────────────────────────────┘

---

02 · OCR Pipeline

---

03 · System Architecture

flowchart TD
    A[PDF / Image Input] --> B[Document Loader]
    B --> C[Text Detection]
    C --> D[Text Recognition]
    D --> E[Structured Output]
    E --> F[Text Export]
    E --> G[PDF Export]
    E --> H[JSON Export]
    I[Streamlit UI] --> B

Loading

---

04 · Key Features

Feature	Purpose
Handwritten OCR	Converts handwritten document content into digital text.
Multi-format input	Supports PDFs and common image formats.
Detection and recognition flow	Separates page analysis from text recognition for cleaner processing.
Structured output	Produces data that can be reused in downstream workflows.
Export utilities	Supports text, JSON, and PDF-oriented output paths.
Streamlit interface	Provides a simple interactive upload and processing experience.

---

05 · AI Workflow

flowchart LR
    A[Document] --> B[Preprocess]
    B --> C[Detect Text Blocks]
    C --> D[Recognize Text]
    D --> E[Assemble Output]
    E --> F[Export]

Loading

Stage	Output
Input loading	PDF or image document converted into processable pages.
Detection	Located text regions and layout blocks.
Recognition	Handwritten content converted into digital text.
Structuring	Text arranged into export-ready payloads.
Export	Text, JSON, or PDF output.

---

06 · Installation

git clone https://github.com/ns7523/ScribbleQ.git
cd ScribbleQ
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

---

07 · Usage

Run the Streamlit app:

streamlit run streamlit_app.py

Open the local interface:

http://localhost:8501

Python API example:

from hwte.io import DocumentFile
from hwte.models import ocr_predictor

model = ocr_predictor(pretrained=True)
doc = DocumentFile.from_pdf("document.pdf")
result = model(doc)

print(result.export())

---

08 · Project Structure

.
├── assets/
│   └── brand/
│       ├── hero.svg
│       └── pipeline.svg
├── demo/
│   ├── app.py
│   └── backend/
│       └── pytorch.py
├── hwte/
├── streamlit_app.py
├── requirements.txt
└── README.md

Suggested production structure:

docs/ · demo/ · hwte/ · tests/ · examples/ · assets/screenshots/ · requirements.txt

---

09 · Visual Assets

Upload Interface assets/screenshots/upload-interface.png Streamlit upload and document selection flow.	OCR Result assets/screenshots/ocr-result.png Recognized text output.
Export View assets/screenshots/export-view.png Text, PDF, and JSON export options.	Pipeline View assets/screenshots/pipeline-view.png Visual overview of detection and recognition stages.

---

10 · Future Improvements

• Add sample input and output examples.
• Add OCR accuracy and latency benchmarks.
• Add screenshots under assets/screenshots/.
• Add deployment notes for Streamlit Cloud.
• Add tests for input loading and export utilities.
• Add documentation for OCR model components.
• Add a clear license file.

---

N S Akash

AI & Cybersecurity Engineer