- Hardware
- Computer Vision
- Full-Stack
Privara
Facial-recognition door access running entirely on a Raspberry Pi 5
Capstone project, Accra Institute of Technology
Project overview
Privara combines computer vision, embedded hardware, backend services, and a mobile interface in a door-access system. Developed as a Computer Engineering capstone at Accra Institute of Technology, it runs the recognition pipeline on a Raspberry Pi 5. Keeping inference on the device avoids sending each recognition request to a remote server and makes the edge device the centre of the access workflow.
The Python and YOLOv8 pipeline combines facial recognition with liveness detection, addressing the difference between recognizing an enrolled face and accepting a static image of that person. A Node.js and Express backend, MongoDB, and JWT authentication support enrollment and access records. The React Native app provides remote monitoring and management, connecting the physical device to an interface that an administrator can use.
The problem to solve
Traditional access-control hardware is either expensive, cloud-dependent, or both, which is a poor fit for a building that needs reliable, low-latency security without recurring cloud costs or a hard network dependency. The brief was to run real-time facial recognition entirely on the edge, with inference on the device rather than round-tripped to a server.
Inside the build
The implementation, from application behaviour to the infrastructure underneath it.
- Run a YOLOv8 facial recognition and liveness-detection pipeline natively on a Raspberry Pi 5
- Defeat spoofing with liveness detection instead of static image matching
- Architect a Node.js and Express backend with MongoDB and JWT authentication for enrollment and access logs
- Pair the device with a React Native app for remote monitoring and management
What shipped
The finished system recognized enrolled faces with over 95% accuracy at sub-second inference latency, entirely on-device, delivered as a capstone project at Accra Institute of Technology.
Built with
- Raspberry Pi 5
- YOLOv8
- Python
- Node.js
- Express
- MongoDB
- React Native
Want to talk through how this fits a system you are building?