Mask detection and alert system
This was a fun and timely project born during the pandemic. One day I saw a sign in my apartment building that read: “Mandatory use of masks in all areas”, and I found myself wondering: “But how would anyone actually check?”
With a Raspberry Pi equipped with a camera sitting idle, I built an end-to-end system to monitor mask usage in real time. If someone was detected without a mask, the system would automatically send a WhatsApp alert.
The pipeline included:
- 🖼️ Building a small custom dataset of masked and unmasked faces
- 🎥 Using a Raspberry Pi camera as a real-time video stream source
- 🧑🔬 Running face detection to crop and extract face regions from each frame
- 🧠 Training a CNN to classify mask usage
- ☁️ Deploying the model to a cloud server with FastAPI for inference
- 📲 Integrating WhatsApp notifications via Twilio
It was a lighthearted side project during uncertain times, and a great opportunity to work with hardware, computer vision, and real-time systems.
🧩 Features
- 🧠 Real-time face detection and mask classification
- 📷 Raspberry Pi as a low-cost edge device
- ☁️ Cloud-hosted inference API for efficient processing
- 🔔 WhatsApp notifications for non-compliance alerts
- 🧪 Custom training pipeline with CNN model
💡 Technologies used
- OpenCV – Face detection and image processing
- TensorFlow / Keras – CNN training for mask detection
- FastAPI – Lightweight inference API
- Raspberry Pi – Real-time video streaming hardware
- Twilio API – WhatsApp messaging integration
- Docker – Reproducible environment for running the server
- Python – Core backend and data flow orchestration