BusyTartan
Edge vision for campus queue estimation
A ScottyLabs Labrador project developing an on-device approach to estimating campus mailroom queues, with aggregate counts rather than a public camera feed.
A practical campus problem
Students need useful information about mailroom queues before making a trip. BusyTartan is being developed to estimate queue length and support future wait-time estimates and historical trends.
Proposed architecture
- A Raspberry Pi 5 captures a frame at approximately 30–60 second intervals.
- A local YOLO-based person detector estimates the number of people in the scene.
- Frames are processed in memory and discarded; the design does not retain video or expose a public stream.
- Only aggregate counts and timestamps are transmitted to a backend for a user-facing dashboard.
This describes the proposed system. Deployment also depends on campus permissions, installation, network security, and device access controls.
Current status & constraints
The project is under development and coordinating with campus stakeholders. It is not presented as a deployed or widely used service. Reliable counting, the relationship between queue length and actual waiting time, and installation constraints remain practical engineering questions.
My specific implementation contributions will be documented as the work progresses.