Acceso restringido

Ingresa el PIN que provee el backend para desbloquear la app.

Road Volumes Estimation

Draw gates, stream counts, and explore traffic flows.

Upload a short clip, drop one or more gate lines, and watch vehicles accumulate by class in real time.

Choose video Model Window (s)

Tip: click two points to create a gate line.

Status: Idle

Waiting for a run.

Total by Gate

Total by Class

Time Series

Accumulated

Trajectories

About Traffic Counter

Real-time vehicle intelligence from pixels to decisions.

Traffic Counter MVP pairs FastAPI job streams with a Vite/Plotly frontend so teams can draw gates, launch jobs, and instantly visualize counts, classes, and trajectories.

~5s to first chart after upload 3 chart surfaces updating in sync 1 cancel tap to stop backend work

Backend streaming

Jobs store the upload/payload under storage/<job_id>, precompute a preview, and expose /api/jobs/{job_id}/stream that emits window, traj, and box chunks as SSE. Automatic cancellation checks and status JSON keep things responsive.

Frontend live charts

EventSource pushes update shared state for windows, trajectories, and latest boxes so Plotly renders gate totals, class totals, time series, accumulated sums, and the trajectory atlas. Controls include the PIN gate, cancel flow, sample upload, and exportable CSV.

Use case wins

Replace manual head counts with automated CSV/data exports, react faster when spikes appear, and give operations and analytics teams a single stream that democratizes both gate-level totals and trajectory detail.

Live dashboard

Everything you see here comes directly from the stream.

Gate totals, class breakdowns, and trajectories update in seconds because each chunk is emitted as soon as the backend detects a crossing. This dashboard is effectively your cockpit: historical CSV exports, instant charts, and the trajectory atlas are all sourced from the same running job, so manual counting and post-processing disappear.

Screenshot of the Traffic Counter dashboard — The dashboard mirrors the live charts and trajectories that the operators use, proving that the streaming backend can feed actionable visuals without waiting for batch work.