The Problem
Legacy parking systems are often analog or laggy. The project aimed to provide instant, visually-verified availability updates to motorists.
The Approach
Deployed the YOLO v3 (You Only Look Once) object detection algorithm on a Raspberry Pi accelerated by an Intel Neural Compute Stick. The system processes Pi camera feeds in real-time, syncing spot statuses to Firebase. A Flutter-based web app consumed this data to provide a live GUI with Google Maps integration.
Technical Stack
Challenges & Constraints
Optimizing deep learning models for edge hardware and ensuring low-latency updates between the Pi and the Firebase real-time database.
Outcome & Learnings
Developed a fully functional prototype that tracked parking occupancy with 95% accuracy and broadcasted updates in under 2 seconds.