Vehicle Registration Recognition
Read out the vehicle registration data with just one photo
Team
3 AI Developers
Project duration
6 months
Product success
More than 200T scans per month by KÜS and other customers
Vehicle registration scanner
​
Together with KÜS DATA GmbH, we have developed an innovative AI product called fahrzeugscheinDIGITAL, which, based on a photo or scan of the vehicle registration, quickly and easily reads all relevant data from a vehicle registration. With more than 200,000 hits per month by insurers, workshop chains, tire manufacturers, testing companies and many more, it is used successfully.
​
What initially started as an internal project became a real joint AI product with fahrzeugscheinDIGITAL.
​
.png)
Our partner:
Our services: Consulting, AI development, deployment, continuous development and support
Technologies used: CNN, RNN, Beam Search, REST API, Docker
"Since we already knew from our experience that good content recognition for vehicle documents is very challenging, it was clear to us that a purely OCR-based solution would not be satisfactory. The implementation of the vehicle registration certificate recognition with the help of deep learning technologies was implemented very well according to our requirements with recognition rates of over 99.5%. In the meantime, this project has even become a successful, independent product with fahrzeugscheinDigital, including a long-term partnership with kaitos.”
Thomas Auer, Managing Director of KÜS DATA GmbH
Reliable
A read rate of over 99.5%
Fast
Two seconds instead of several minutes
Secure
The data is processed GDPR-compliant under the highest standards
And how does it work?
No matter which device, no matter from which angle: simply take a picture of your vehicle registration document
Take a photo or scan
The image is uploaded to the KÜS DATA web service via a REST-API interface. The data processing is GDPR-compliant.
Upload image to KÜS API
The KI-OCR scans the image file for several patterns, such as vehicle identification number, date of first registration, HSN/TSN etc. Incorrect results are detected using the KI's plausibility check.
Image is processed by AI service
After the data has been processed by the KI-OCR, it is sent back in JSON format. For each field scanned, the API returns the five most likely values and their detection probability.
Output of files in JSON format
For more information
or information about our services
.png)