Fast · Modular · Real-Time Facial Recognition Attendance System
Built with Django, OpenCV, and Deep Learning–Based Face Encodings
ReconRoll is a modular, real-time facial recognition attendance system designed for automation in schools, workplaces, events, and controlled environments. It provides robust attendance tracking using fast face detection, deep-learning encodings, and session-based logging.
ReconRoll evolved from the earlier FaceTrack Lite project and demonstrates a practical application of computer vision and AI engineering.
- Real-time face detection & recognition
- Automated attendance logging
- Capture of unknown faces for later review
- Attendance sessions with summaries
- Offline-first operation
- Lightweight, simple admin interface
| Technology | Purpose |
|---|---|
| Python | Core backend language |
| OpenCV | Face detection + video processing |
| face_recognition | Deep learning face encodings (dlib) |
| Django | Backend logic and web framework |
| SQLite/PostgreSQL | Database options |
| Docker | Containerized deployment |
| Bootstrap + JS | Admin UI and interactivity |
ReconRoll is designed around four core subsystems:
- Performs face detection and encoding
- Compares encodings with enrolled users
- Optimized for fast recognition using OpenCV pipelines
- Starts, runs, and ends attendance sessions
- Logs recognized users in real time
- Stores unidentified face captures
- Handles user registration
- Captures and processes face images
- Generates stable 128D embeddings for matching
-
Databases for:
- Encoded faces
- Attendance records
- Unknown captures
- Session histories
-
Defaults to SQLite but supports PostgreSQL
Admin uploads/captures user photo → generates encoding → saved in DB.
Admin begins a session → system starts processing frames.
Frame-by-frame:
- Face detected
- Face encoded
- Recognized → logged
- Unknown → saved
ReconRoll produces:
- Present students
- Absent students
- Unknown face log
- Attendance summary
| Module | Role |
|---|---|
| Recognition Engine | Detects, encodes, and matches faces |
| Enrollment Service | Registers new users and creates embeddings |
| Session Manager | Controls session lifecycle and logging |
| Unknown Face Handler | Stores unknown captures for review |
| Analytics Engine (Planned) | Attendance stats, performance metrics |
Before running, you may review the Pre-Installation Guide: https://docs.google.com/document/d/1OgYudT0YOkN6vht0wn9dWe4mxkAFOjGnz5ulX36hd94/edit?usp=sharing
git clone https://github.com/peter-njoro/ReconRoll.git
cd ReconRoll# Linux
chmod +x run.sh
./run.shApp available at:
http://localhost:8000
git clone https://github.com/peter-njoro/ReconRoll.git
cd ReconRoll
python -m venv venv
source venv/bin/activate # Linux/macOS
venv\Scripts\activate # Windows
pip install -r requirements.txt
python manage.py migrate
python manage.py runserver- Camera captures input
- Face detected
- Face recognized or marked unknown
- Attendance logged instantly
Contributions are welcome!
- Fork the repo
- Create a feature branch
- Submit a PR to development
ReconRoll is designed for educational and demo purposes. It is not ready for high-security or large-scale deployments.
Peter Njoroge Chege Machine Learning Engineer (In Progress) AI • Computer Vision • Backend Engineering
Inspired by the original Virone concept by Everlyne Mwangi.
If you’re reading this part:
- Yes, pip errors still haunt me.
- Docker promised peace, webcams declared war.
- ReconRoll is both a demo and a flex.
- And yes… by the way… I use Arch btw .
- If this becomes Skynet, at least the README will survive.