About
lpr.software is built by DeepVision — a computer-vision team that trains, deploys, and operates license plate recognition in production, not a wrapper around someone else’s model.
97%+
Plate read accuracy
<200ms
Typical response time
99.9%
Uptime SLA
2,003
Car models recognized
DeepVision started as a computer-vision team solving a concrete problem: reading license plates reliably in real parking facilities, where cameras are imperfect, weather is uncooperative, and every misread costs an operator money. Our models are trained on real-world traffic imagery — vehicles at angles, in motion, at night — not curated benchmark datasets.
Our home market is Kazakhstan, where the system runs live parking operations today, processing real traffic around the clock. Running our own models in production keeps us honest: we see every edge case, retrain on new data, and fix what breaks — because we are the ones on call when it does.
Now we are opening the same engine to the world as a self-serve API. No sales calls, no pilots that take a quarter to start — an API key, 150 free calls a month, and documentation that respects your time. The hard Latin-script problems are our focus: 50 US state plate designs, stacked characters, and vanity plates, alongside Canada, the UK, Australia, and New Zealand.
Philosophy
Every field the API returns carries a confidence score, and we would rather say nothing than say something wrong. Our brand and model recognition runs two independent networks with calibrated confidence gates — when they disagree and the disagreement cannot be resolved, the API returns null with an explicit uncertainty flag instead of a plausible-sounding guess.
That design philosophy runs through the whole product: plate reads you can threshold, vehicle attributes you can cross-check, and errors that announce themselves instead of hiding in your data.
{
"make_model": null,
"brand_uncertain": true,
"decision": "disagreement_unresolved"
}lpr.software is a product of DeepVision, a computer-vision company headquartered in Kazakhstan. We build and operate recognition models for license plates, vehicle brands and models, body types, and colors — available as a cloud API or deployed on-premise in your infrastructure.
Send us sample photos from your cameras and we will run them through the pipeline — free, with the raw JSON included.