TrackStore

• 

  Logo

• 

  Dashboard with data (manipulated in supabase)

• 

  Website (un published design on Figma)

• 

  Website (un published design on Figma)

Inspiration

My Unkle owns several Shops, and sometimes i asked him, hey, how much traffic do you actually have ? What are your peak hours etc. He couldnt answer, without looking throught the amounts of reciepts he got that day. Only because of this i came onto the idea to automate this and build a Store tracker.

What it does

It uses YOLO and a RaspberryPI 5 to track and process frames, if people are in those frames, they will get recognized by the YOLO model, and they get added to a counter and statistics. The device will count if people enter/leave the store, and the time they are staying in the store. You can also set a Maximal number of people who can go in the store, and it will show you, if this limit is hit.

How we build it

My friend built the computer-vision part — the YOLO model that detects and tracks people and turns line-crossings into enter/exit events. I built the frontend: the live dashboard, the charts, the real-time updates from Supabase, and the settings and reporting features. Supabase ties it together — the tracker writes events, and the dashboard updates live over websockets. We also used a bit of AI to help with the store-tracking part.

(Note: my teammate couldn't make it into the video, but the tracking side is his work.

Accomplishments that we're proud of

It actually works end to end: a phone camera counts real people, and the numbers show up on the dashboard live, within a second. We also built extras on top — dwell-time analytics, an AI guide that answers questions about the store, and an automated weekly email report.

What we learned

How object tracking really works (detection vs. keeping a consistent ID), how to build a real-time app with Supabase without polling, and how to draw charts from scratch on a canvas. Mostly, how to split the work and get two very different pieces — vision and web — to talk to each other cleanly.

What's next for TrackStore

Multi-camera support for bigger stores, historical trends (week/month/year), heatmaps of where people spend time, and turning the weekly report into full automated insights and alerts.

Built With

• html
• javascript
• pythion
• resend
• supabase
• typescript