By using AWS re:Post, you agree to the AWS re:Post Terms of Use
AWS re:Postre:Post

Extract details from an ID (Not US based ID)

0

Hi, I am currently exploring AWS services for a project (KYC Processing) and I would like to know if I can extract data from an ID (Ex: Driver License). Textract already provide this service but it only allow US Driver License and Passport. Is there any way I can do so that it can also cater driver license from other country?

WHAT I KNOW:

  1. Use Textract Analyze ID
  2. Use AWS Rekognition text to image service (It can be used, but what I want to do is return a key-value pair for the details and the service only returns the text)

QUESTIONS:

  1. For #1, It only supports US Driver License and Passport. Is there any way I can customize it so it could also read driver license from different country
  2. For #2, I can create a custom label, train my model, and it can identify the details from the ID. Is there any way I can also extract it, not just provide a bounding box on it?
  3. If I am going to use a custom label, what is the required quantity of the dataset to make my custom model accurate? Should I use different ID's of different persons or a multiple image of the same ID but with different cases (Different Angle, Different lighting, or Different position on the image).
Topics
Machine Learning & AI
Tags
Amazon TextractAmazon Rekognition
Language
English
tbd-dave
asked a year ago829 views
1 Answer
0
Accepted Answer

In order to extract the data for a non-US ID or Driving license you can use Textract Query https://docs.aws.amazon.com/textract/latest/dg/API_Query.html

Please check this example for the UK (dummy sample), Enter image description here

as you can see in the example, I used "what is the DATE in 3." or "what is the COUNTRY in 3." to extract the correct subfield in the row main field "3."

Using this feature of Textract together with Rekognition Custom Labels to recognize the type & country of the ID , you can leverage a Step Function logic flow for your problem. Note that Rekognition Text detection is mostly used for road signs, logos and not for document intelligent OCR (Optical Character Recognition) extraction, therefore Textract is a best fit.

[Please accept my answer if you consider it helped you]

profile pictureAWS
Alexandru Raileanu
answered a year ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions

Relevant content