Sighthound Retriever - Advanced ALPR and Vehicle Recognition
A browser application that reads license plates and identifies vehicles by make, model, colour and generation. Runs on your hardware, in your environment.
Live camera streams or uploaded footage. No cloud connection required.
What happens when the plate doesn't read?
Dirt on a plate. Rain. A bad angle at the gate. A tow bar sitting across the numbers.
Most of the time the read comes back clean. When it doesn't, something still has to happen: someone checks the camera feed, looks up a record by hand, or waves the vehicle through and reconciles it later.
Every operator we've spoken with has an answer to this. Most of those answers involve a person.
The cost isn't the failure rate. It's keeping someone available for a case you'd describe as rare.
A plate is one signal. Retriever returns four more.
Alongside the plate, Retriever returns the vehicle's make, model, colour and generation.
That matters most in the moment the plate itself isn't usable. A partial read plus a silver 2019-2022 Cadillac CT5 is often enough for the person reviewing it to resolve the exception without pulling footage.
It also means your search isn't limited to plates you already know. You can look for a vehicle by what it is, not only by what's on the back of it.
Runs on your hardware. Stays in your environment.
Retriever installs on your own machines through a Docker Compose stack and runs in the browser. It works offline. There's no required cloud connection and no dependency on us being reachable.
Your media never leaves your environment, compliance stays under your policies and controls.
A dashboard your team can actually run
No SSH sessions, no config files to hand-edit. The things an operator needs during a shift are in the browser.
| Item | What it does |
|---|---|
|
Live dashboards
|
Watch active camera streams and detections as they happen. |
|
Upload processing
|
Drop in recorded footage and search what it finds. |
|
Regions of interest
|
Draw the zones that matter on a live stream and ignore the rest. |
|
Hotlist matching
|
Flag plates on a watchlist the moment they’re seen. |
|
Search
|
Find a vehicle by plate, make, model, colour, or generation. |
|
Analytics
|
See what came through and when it came through. |
See it run on your own image
Upload a photo and Retriever returns the plate, plus make, model, colour and generation. No call, no procurement conversation.
The demo shows the identification. The full response, every field, every confidence value, is available in a technical fit review or through Retriever API.
Retriever API - for developers and integrators
Building vehicle recognition into your own platform? Retriever API returns the same plate and vehicle information programmatically.
Is Retriever a fit for your environment?
Tell us what you're running: live streams, uploaded footage, or both. What happens today when a plate can't be confirmed, and the deployment constraint you're working around?
We'll tell you what fits Retriever today, what still needs engineering confirmation, and whether a review makes sense. If it doesn't, we'll say so.