StreetFlow AI

Turn your phone into a smart, on-device traffic counter. This app uses your camera and local AI to detect and count people, bicycles, and vehicles in real time (no sign in, no cloud processing).

What it does

Live detection & counting: Tracks people, bikes, and vehicles in the camera view.

Counting line: Draw a virtual line; items are counted once as they cross it—reduces double-counts.

Adjustable sensitivity: Tune detection strictness for different scenes and lighting.

Front/back camera support: Switch between the rear (“environment”) and front (“selfie”) camera.

Real-time stats: See totals plus per-minute and per-hour rates.

Session insights: View session start time, active detection time, and last update.

Data export: One tap to download your session data as JSON for analysis.

Clean UI overlays: Full-screen video with compact controls and a collapsible stats panel.

How to use

Tap Start to begin detection.

Tap Line and set start/end points to define where counting occurs.

(Optional) Open Settings (gear icon) to adjust sensitivity or switch cameras.

Watch the live counters update and export results when you’re done.

Works entirely on your device

All processing runs locally on your phone using an embedded model. No images or video are sent to a server. Exported JSON contains only counts and session stats.

Great for

Quick traffic counts (people flow, cycling activity, road vehicles)

Event operations and queue monitoring

Campus, retail, and venue footfall snapshots

Small studies and rapid field assessments

Tips for best results

Place the counting line where subjects naturally pass (e.g., curb edge, doorway).

Use the Back camera for wider FOV and steadier framing.

Increase Sensitivity in low-contrast scenes; decrease it to reduce false positives.

Ensure adequate lighting and keep the device steady.

Permissions

Camera: Required for detection. No media is stored unless you export data yourself.

Disclaimer

This tool is intended for informational use only and should not be relied upon for safety-critical applications. Accuracy may vary with environment, motion, lighting, and camera quality.