ChargePoint - Real-Time Camera Recognition of People and Vehicles using OpenCV + YOLO
ChargePoint was looking to camera-based technology to solve a few problems: parking space occupancy (in the case a vehicle had parked but not plugged in), people detection (allowing for power-saving or audio mute when no audience), and ICE-ing (where the parking spot is blocked by non-EV vehicles).
As I had previous experience with OpenCV and Haar-Cascade based facial recognition at Skype (to switch between local users using the camera on a Smart TV), I was given the opportunity to build this for ChargePoint.
YOLO (You only look once) is a massive improvement on previous basic object recognition, so much so that it allows for Real-Time recognition on embedded ARM based platforms. Utilizing simultaneous detection of people and vehicles, as well as identifying parking regions, allowed for both parking occupancy detection and active audience detection.
Unfortunately as European rules around personal data and images increased (ie. GPDR), and interest in cameras on public kiosks waned, this didn’t end up being productized.
