Car image used for testing
Super simple hardware setup!
What it does
Proof of concept for a smart dashcam. It takes a photo every few seconds and then uploads it to the Google cloud platform, where my teammates' machine learning code extracts the licence plate and car colour, which is then ran against the DVLA's records to check for discrepancies, allowing stolen vehicles to be identified easily.
How I built it
Generally plug and play and a few open source libraries such as the Google Cloud SDK and the HTTP request library. Also relies on my phone as a hotspot. A bit of SSH involved and a lot of learning how to actually use a Pi
Challenges I ran into
Learning the proper Raspberry Pi toolchain/installing the right dependencies. Learning the basics of SSH and RDP and how to deploy software on a monitorless Pi.
Accomplishments that I'm proud of
My first ever Hackathon contribution - and it worked!
What I learned
How to make effective use of a Raspberry Pi, and a basic command of Python, especially useful libraries for interacting with web APIs.
What's next for piWebcamANPR
Perhaps using the stock RasPi camera for better resolution, and an image cache for when the device is offline. This has only been tested on a big smart board with images of cars, and a live test in a real vehicle is definitely necessary to determine things like optimal camera angle and lens zoom. Has a lot of potential for accident detection and providing good data for telemetry insurance policies.