MarkMyPaper is a Flask-based web application for automated grading of handwritten or typed answer submissions. It combines OCR (TrOCR), PDF/image text extraction, and fuzzy text matching to compare student responses against model answers.

• User authentication (register, login, profile, logout)
• JWT-protected API routes
• Password reset token flow
• Upload support for PDF, JPG, JPEG, and PNG
• OCR/text extraction pipeline:
  • TrOCR for images
  • PyPDF2 for PDFs
• Weighted answer grading using similarity scoring
• Result breakdown per model answer + total score

• Python 3
• Flask, Flask-SQLAlchemy, Flask-CORS
• Transformers (microsoft/trocr-base-handwritten)
• PyTorch
• Pillow, PyPDF2
• TheFuzz (token_sort_ratio) for grading similarity
• SQLite (via SQLAlchemy)

MarkMyPaper/
├── app.py
├── requirements.txt
├── templates/
│   ├── dashboard.html
│   ├── home.html
│   ├── login.html
│   ├── trial.html
│   └── upload.html
└── static/
        ├── home.css
        ├── login.css
        └── styles.css

git clone https://github.com/DysCreate/MarkMyPaper.git
cd MarkMyPaper

python -m venv .venv

Windows (PowerShell):

.\.venv\Scripts\Activate.ps1

macOS/Linux:

source .venv/bin/activate

pip install -r requirements.txt

python app.py

The app starts in debug mode at http://127.0.0.1:5000.

On first run, Hugging Face model files for TrOCR are downloaded, so startup may take longer.

• POST /api/register
• POST /api/login
• GET /api/user (requires Authorization: Bearer <token>)
• POST /api/logout (requires token)
• POST /api/reset-password-request
• POST /api/reset-password

• POST /upload
  • Form-data fields:
    • file (PDF/image)
    • answers (repeatable)
    • weights (repeatable)
  • Returns extracted text, per-answer similarity/score, and total score

• User data is stored in local SQLite (instance/users.db).
• SECRET_KEY is currently hardcoded in app.py; for production use, move it to an environment variable.
• This project is set up for local development and experimentation.

This project is licensed under the MIT License. See LICENSE for details.