Bicyclist Defense

Florida has one the highest death rates for pedestrians and cyclists in the United States.

What if cyclists could have situational awareness for erratic drivers at all times?

We aim to train a mmWave radar and computer vision system to alert bicyclists to erratic drivers, before an accident.

The Jetson Orin Nano, Intel Realsense, and AWR1443 (via 5v BEC) are all powered by a DeWalt 20v battery.

The project consists of a FastAPI web application with the NanoOWL model running on the Jetson Orin Nano, connected to the internet via the hotspot of the bicyclist's iPhone.

The web application is reachable from the iPhone via a Tailscale network, and drives a real time video stream of cars behind the cyclist on the iPhone.

Vehicles and pedestrians are highlighted in the video stream.

While riding, the cyclist has the option to record a video of their ride (saved as H264 to disk with a timestamp of the beginning of the ride), or to save mmWave radar data alongside the recording as a rerun file for later analysis / model training.

The goal is to enhance bicyclist safety, and build a warning system for cyclists.

Code is available at Github.