RadarHub watches a video feed, detects every vehicle, measures its real speed in km/h, counts traffic by type, and flags speeders with photo evidence. Live web dashboard included. Runs on your own hardware: no cloud, no per-camera fees, no subscription.
Real pipeline output: green under the limit, red flagged at 128 km/h, live counts top right.
The same pipeline used by professional enforcement systems, packaged so one person can run it.
Pretrained YOLO models detect cars, trucks, buses and motorcycles out of the box. No training, no labeling, no dataset work.
A calibrated perspective transform maps camera pixels to road meters. Speed is measured on the ground plane, the defensible way.
Every vehicle keeps one stable ID across frames, so speeds, counts and violations always belong to the right vehicle.
Every speeder is logged with a cropped snapshot, its speed, class, and timestamp. Export the full log to CSV in one click.
A virtual line counts vehicles by type as they pass. Perfect for traffic studies, parking flow, and site planning.
Watch the annotated feed, live counters, per-vehicle speeds and the violations feed from any browser on your network.
Point it at a video file, a live RTSP/IP camera, or a webcam / OBS virtual camera. Switch sources from the dashboard.
Click 4 points on the road, enter their real size in meters, done. Calibration is saved per camera for accurate speed forever.
Python source code included. Your footage never leaves your machine. GDPR-friendly by architecture, white-label ready.
One command: pip install -r requirements.txt. Works on Windows, macOS and Linux. GPU optional.
Click 4 points on a road patch you can measure (one lane over 20 m works). This turns pixels into meters, and meters into km/h.
Open the dashboard, point RadarHub at your camera, and see live speeds, counts and violations with photo evidence.
Cloud vision APIs bill you per camera, per month, forever. RadarHub is a one-time purchase that runs on hardware you already own.
| RadarHub | Cloud vision SaaS | Hardware radar unit | |
|---|---|---|---|
| Price model | One-time | €99+ /camera /month | €2,000+ per unit |
| Works with existing cameras | |||
| Footage stays on-premise | |||
| Speed measurement | Rarely | ||
| Vehicle counting by type | |||
| Source code included | |||
| White-label / resell | Extended license |
Monitor residential streets, school zones and campus roads. Collect evidence-grade data before investing in enforcement.
Count gate traffic by vehicle type, enforce site speed limits, and export daily CSV reports for management.
Run traffic studies from any camera footage. The Extended license lets you deliver RadarHub-powered reports under your own brand.
No. On CPU the default nano model processes video smoothly for analysis work. For real-time processing of high-resolution live streams, any consumer NVIDIA GPU (even a GTX 1650) gives a large speedup.
Accuracy depends on calibration quality. With honest measurements of the reference rectangle (for example a lane width and a marked distance), real-world accuracy within a few km/h is achievable. The docs include a full calibration guide.
RadarHub is a measurement and analytics tool. Official traffic fines usually require type-approved devices in most countries. RadarHub is ideal for studies, private sites, evidence gathering and monitoring; check local regulations for enforcement use.
Any camera that produces a video file, an RTSP/HTTP stream (virtually every IP/CCTV camera made in the last decade), or shows up as a webcam, including OBS Virtual Camera.
Never. Detection runs locally, the dashboard is served from your own machine, and violation snapshots are stored on your own disk.
License-plate recognition on violation snapshots, multi-lane directional counting, and alerting (email, webhook, Telegram) are planned as add-ons.
One-time purchase, full source code, lifetime license.
Get RadarHub