Computer vision-based process automation

“Professionalism meets passion – the project was implemented very quickly and professionally.”
Torsten Froese, Head of Strategic Innovation & digitization, Bank11 für Privatkunden und Handel GmbH


BANK11’s business involved the daily, repeated typing in of data from vehicle-related documents. In total, BANK11’s employees were processing around 190,000 vehicle registration documents per year – by hand. The large number of manual operations involved was not only time-consuming but also costly and prone to error. It also represented a burden for the company’s employees, keeping them from a much more important task: that of providing a good service and to quickly approving credit applications.

Bank11 was therefore eager to automate the processes involved. Although standard software solutions exist for reading documents, the extraction of data from vehicle registration documents represents a special case that is subject to specific stumbling blocks.


Bank11 is a credit institution that specializes in sales financing. It focuses on assisting smaller and medium-sized vehicle dealerships with competitive mobility services and insurance products. Bank11 provides straightforward and cost-effective finance products for end customers and robust liquidity solutions for automobile dealerships. Thanks to the use of digital “floor checks,” Bank11 also offers significant efficiency gains in the purchase of financing, thereby saving time and reducing risk. Bank11’s range of services is also distinguished by the skilled personal support provided by a specially trained field and office-based team.

Our solution

As part of a proof of concept, we read the VINs (vehicle identification numbers) from vehicle registration documents. In addition to the difficulties related to reliably recognizing where the VIN was located within each document, the different texture of the various vehicle documents posed a particular challenge. It was therefore necessary to “tidy up” the different fonts, colors, and print intensities involved, printed matter that went outside the form lines, and documents that had been scanned upside down. This was done with the aid of a technique known as “scan preprocessing.” With the help of automatic image processing operations, the scans became clearer and more legible, ensuring that they could ultimately be recorded in a standardized way. As a result, the system is now successfully reading the VINs with a 92.31% rate of accuracy. Before our tools were deployed, the accuracy rate was less than 20%, meaning that the automated solution could not be used.


In the future, it should be possible to read all of the data from the vehicle registration documents automatically in order to avoid complicated and error-prone data input by hand. This will save money and time while reducing the employees’ workload. The data from the vehicle documents can also be used for further innovative optimizations. For example, customers might automatically receive information and the loan approval for a vehicle on their device by means of a smartphone scan. This would clearly enhance the customer experience.


In addition to the automation workflows, the data collected could be usefully processed as a basis for further analysis. By adopting creative and innovative approaches and advances in digitization, Bank11 can use such forms of data analysis and processing to open up a large number of further opportunities in the future.

Key Facts

  • Increase in accuracy: from 18% to 92.31%
  • Application of deep-learning OCR tools
  • Complex preprocessing tailored to vehicle registration documents
  • Development of a separate validation procedure for results in the OCR (optical character recognition) field


  • Tesseract
  • OpenCV’s EAST
  • Google Vision API
  • Python
  • Neural Networks


  • Agile project management
  • Kanban

About the project

Torsten Froese
Head of Strategic Innovations & Digitization
BANK11 für Privatkunden und Handel GmbH

As part of our proof of concept, we worked with tarent on the goal of automating the way we capture vehicle registration documents (authorization certificates) and enhancing data quality. The result? We’ve been able to reduce processing time from 1-2 minutes to 30 seconds and the error rate from around 20% to 5% – with a total volume of approx. 100,000 letters. All in all, the colleagues from tarent implemented the project very quickly and professionally – and showed real passion for their work.

Marissa Elble
Data Scientist
tarent solutions GmbH

Working on this exciting OCR project for Bank11 was an awesome experience. It was a great challenge and one we were able to successfully resolve with creativity and the latest technologies. The constructive and straightforward cooperation with Bank11 was particularly positive for us. We are very proud of the sensational result, which exceeded all of our expectations.

Technology and experience

tarent in figures


years of experience


office locations

300 +

IT professionals


digitization experts




parallel development projects


    Then get in touch.

    Request a no-obligation fact-finding appointment now.