Model AI OCR Read & Filter Insurance card Information

A large corporation in Japan in the field of providing business operation services, accounting – finance & real estate management.
Model AI OCR Read & Filter Insurance card Information

Challenges

The customer assigns about 200 employees the task of data entry – which is high cost, time consuming & there is still a risk of Human errors in the data entry process.

There are 2 types of documents with a large amount of information to enter: Insurance Card & Bill. They need to find a faster & more cost-effective method for this process.

HBLAB's Solutions

Combine multiple AI Models to create a total solution:

  1. Use Image Processing to re-edit the image (shooting angle, photography direction, blur, shadow)
  2. Use Object Detect to catch the area containing the information needed
  3. Read information:
  • Insurance card: Use OCR to read
  • Bill: Use OCR to read both print & handwriting. Use Logo searching to collect & classify data about suppliers information.

Customers demand to collect data that change over time and for different work purposes. As a result, HBLAB proposed different AI models to solve changing problems for customers

Project details

  • Used Technologies Image Processing, Object Detect, OCR, DewarpNet, Text DetectionLogo Searching
  • Development Team 5 AI Engineers + 0.5 PM + 0.5 Comtor
  • Duration 1 year

Results

The logo & store name accuracy is 93%, the other information accuracy is 95%-97%.

The AI ​​model built by HBLAB helps to save 20% of the customer’s initial data entry, optimize costs & speed up work.

Related cases

Explore our success stories to spark ideas for your business and discover why partnering with us is the right choice.
A company in Japan specializing in smart electronic devices for pets
A startup company based in Japan
The client, which is a large Singaporean IT provider, operates an IT issues ticket/request system that allow users to manage equipment, other users, and handle, create, process, and track IT incident requests within the company
Scroll to Top