Hen-Tersection 🐔

Inspiration

I got tired of waiting at traffic lights with no cars around. Most intersections still run on fixed timers that don't actually know what's happening on the road, so you end up standing there for no reason. I wanted to build something that could actually react to the situation in real time.

What it does

Hen-Tersection is an AI-powered crosswalk system that uses computer vision to manage traffic signals based on what's actually happening at the intersection.

• Detects jaywalkers entering the crosswalk during a green light and triggers a yellow to slow cars down
• Extends the walk signal automatically when it detects someone who needs more time to cross
• Logs every event to Snowflake so the data can be reviewed and analyzed later

How we built it

We built a CV pipeline to process a camera feed and detect pedestrians and crosswalk activity. That feeds into a traffic light controller that handles the phase logic — when to go yellow, when to extend a walk signal, etc. Events get streamed to Snowflake as they happen. The demo frontend is an animated intersection that shows the whole detection and response loop in action.

Challenges we ran into

CV and Snowflake integration were the two biggest ones. Getting the model to reliably detect what it needed to without too many false positives took a lot of tuning. On the Snowflake side, figuring out how to structure and write live event data cleanly was more involved than expected.

Accomplishments that we're proud of

The pedestrian timing logic. It handles a lot of overlapping conditions — a jaywalker mid-crossing, a car-green phase, someone who needs extra time — and getting all of that working together correctly was the hardest part of the project.

What we learned

When software decisions have real safety consequences, you have to think a lot harder about edge cases. We also got solid experience using Snowflake as a live event store, which worked better for this use case than we expected going in.

What's next for Hen-Tersection

Scaling to multiple intersections and using the data to optimize traffic flow across a whole city. Instead of just reacting at one crosswalk, the system could track where cars are building up and adjust signals network-wide to keep things moving.

Built With

• cv
• html5
• python
• snowflake
• yolo