Privasee

A lightweight, privacy-first application that automatically detects and censors sensitive information in images.

Overview

It uses computer vision and NLP techniques to blur or pixelate:

• Human faces
• License plates
• Personally Identifiable Information (PII) such as names, dates, card/account numbers

All processing happens locally on-device — no images are ever uploaded externally. Original uncensored files are stored securely and protected with a password system, ensuring that only authenticated users can view them.

---

Features

• Automatic face & license plate detection (OpenCV Haar cascades)
• PII detection with OCR (Tesseract) + NLP (spaCy) + regex patterns
• Optional Presidio Analyzer integration for enhanced PII recognition
• Choice of Mosaic pixelation or Gaussian blur, with adjustable strength
• Password-protected viewer for originals
• Batch processing of all images in the images/ folder
• Local-first privacy – no cloud uploads

---

Tech Stack

• Language: Python 3.10+

• UI Framework: Streamlit

• Libraries & Tools:

  • OpenCV – face & license plate detection
  • Tesseract OCR – text extraction
  • spaCy – Named Entity Recognition (NER)
  • Presidio Analyzer – advanced PII detection (optional)
  • NumPy – array manipulation
  • Hashlib & JSON – password management

---

Installation & Usage

1) Clone the repo

git clone https://github.com/JordanTwz/tiktok-techjam.git
cd tiktok-techjam

2) Install dependencies

pip install -r requirements.txt

3) Install Tesseract OCR

Windows: https://github.com/UB-Mannheim/tesseract/wiki

Linux (Ubuntu/Debian):

sudo apt-get install tesseract-ocr

macOS (Homebrew):

brew install tesseract

Note: Update the path in backend.py and face_plate_censor_app.py if Tesseract is installed elsewhere.

4) Run the app

Password-protected app:

streamlit run frontend.py

Lightweight demo (no password system):

streamlit run face_plate_censor_app.py

5) Workflow

1. Set password on first use
2. Place images in the images/ folder
3. Censored outputs are saved in the censored/ folder
4. Authenticate with your password to view original images