How to install Tolling Vision
Set up Tolling Vision in minutes: pull the Docker image, add your license key, and connect to our gRPC API for instant ANPR, MMR &...
Our innovative Tolling Vision software is specifically designed for tolling and image review companies, utilizing
Tolling Vision not only identifies vehicle details but also flags vehicles carrying hazardous materials, ensuring comprehensive and safe tolling solutions.
Contact us today to learn how we can streamline your operations and maximize efficiency.
Curious about what our cutting-edge software can do? Click on any demo and witness the power of our technology firsthand. Whether it’s license plate recognition, vehicle make and model identification, or hazardous goods detection, experience the precision and speed of Tolling Vision in real-time.
Don’t just take our word for it—see it with your own eyes. Try a demo now!
Set up Tolling Vision in minutes: pull the Docker image, add your license key, and connect to our gRPC API for instant ANPR, MMR &...
Learn to integrate Tolling Vision via gRPC, call the ANPR, MMR & ADR APIs, and optimise your workflow with best-practice tips—all in under 15 minutes.
Tolling Vision is an AI-powered, Dockerized server that recognises licence plates (ANPR/ALPR), vehicle make & model (MMR) and hazardous-goods placards (ADR) in still images or image sequences, returning structured data in real time.
We release an updated Docker image roughly once per quarter. Each image bundles the latest ANPR / MMR / ADR models together with their engine code, so you must pull the new image tag and redeploy to benefit from the improvements.
Because the upgrade is not automatic, you can (and should) run the new container in a staging environment first, compare its output against your own regression set (specific vehicle classes, plate formats and camera angles), and promote it to production only after you’re satisfied that no accuracy has regressed.
Independent tests of the underlying plate- and vehicle-recognition core show:
Licence-plate (ANPR/ALPR): 95 – 99 % real-life accuracy across diverse countries, and > 99 % in controlled lab datasets; some toll deployments report up to 99.9 % precision when camera optics and lighting are fully optimised.
Vehicle make-and-model (MMR): up to 98 % top-1 accuracy under ideal conditions, with robust performance in low-light or motion-blur scenes thanks to training on millions of labelled images covering 240+ makes and 1,700+ models.
Your own results will depend on camera resolution, angle, and image quality, so we recommend benchmarking the API against a representative sample of your traffic before going live.
Choose between credit-based plans (from 1 M to 50 M recognitions/month) or unlimited, thread-based plans for 24 / 7, high-frequency lanes. Pay-as-you-go top-ups are also available.
Submit JPEG, PNG or BMP stills—or extract frames from video and send them individually. A single request may bundle multiple views (front, rear, overview) and counts as one credit per vehicle. For reliable results aim for ≥ 720 p images.
Every account gets 100 free credits, step-by-step tutorials, and open-source sample clients. Need more help? Our engineers answer questions via e-mail or phone within one business day.
All traffic travels over gRPC with SSL/TLS. You may terminate TLS at a load-balancer or provide your own cert/key pair inside the container; mutual-TLS is supported for client auth.
Engines run locally, but the licence validator calls home for a few hundred bytes per processed image. Short outages are cached, yet fully offline operation isn’t supported.
If the backlog exceeds its limit (Backlog
, default 10) the API returns RESOURCE_EXHAUSTED
; if processing exceeds RequestTimeout
(default 30 s) it returns DEADLINE_EXCEEDED
. Tune those env-vars or scale threads to avoid drops.
Use one of our pre-generated SDKs (Java, Python, Node) or generate your own from the .proto
file; communication is pure gRPC so any language that supports Protocol Buffers works.
Yes. The Docker image is published for both x86-64
and arm64
; simply set ARCH
when pulling the image. Performance scales with available CPU / RAM (≈ 3 GB & 1 vCPU per processing thread).
We maintain open-source clients for Java, Node.js and Python—each on GitHub under smartcloudsol/tollingvision-<language>-sample
. Use them as drop-in templates or reference the .proto
file to generate code in any gRPC-aware language.
We utilize AI-based ANPR/MMR engines that are continuously maintained and regularly updated with the latest license plates and vehicle models, ensuring the highest level of precision and reliability in license plate and vehicle recognition.
Our application is designed to provide secure and fast communication, guaranteeing a seamless and efficient user experience every time you use our services.
Our solutions are built for easy integration with your existing systems, allowing for a smooth and hassle-free setup. Enjoy the benefits of our advanced technology with minimal disruption to your operations.
With extensive experience in software development, system integration, and license plate and vehicle recognition, we are well-equipped to deliver top-notch solutions for you.
Our expertise ensures that integrating our system is straightforward, leading to significant efficiency gains for your operations.
Trust us to provide reliable and precise license plate and vehicle recognition services using continuously maintained and regularly updated ANPR/MMR engines, enhancing your business performance.