Extract Data from an Image Using AWS Textract

Overview

Modern technology has solved this problem to a large extent and data can be extracted from structured forms without human touch. In other cases, however, data is received from a wide variety of unstructured documents without any rhyme or reason to the way the information is presented. Many businesses and government organizations extract data manually from scanned documents, such as PDFs, tables, and forms, which are slow, expensive, and prone to errors. Textract uses machine learning to handle any type of document in real-time, accurately extracting text, forms, and tables without any specification and code.

About AWS Textract

Amazon Textract is a highly scalable machine learning (ML) service that automatically extracts text, handwriting, and data from documents like images, pdf, etc. It can also analyze a document such as related text, tables, key-value pairs, and selection elements. Use Amazon Textract to detect and extract text in your documents.

When the Amazon Textract operation processes the document, the results are returned in an array of Block objects or an array of Expense Document objects. Both objects contain information that has been found about items, including their location in the document and their relationship to other items in the document.

Use Cases

• Import documents and forms into business applications
• Creating smart search indexes
• Creating automated workflows for document processing
• Maintaining compliance in document archives
• Text Extraction for Natural Language Processing (NLP)
• Text extraction for document classification

Architecture Diagram

Steps to Setup AWS S3

Step 1: Open AWS S3 Console

Step 2: Click on Create Bucket. Enter the bucket name (i.e., data-extract-from-image) and select the region that you want to perform.

Step 3: Click on Create Bucket.

Steps to Setup Amazon Lambda

Step 1: Open Aws lambda console.

Step 2: Click on create function and enter the function name (i.e., textract-lambda). Then select the python 3.9 version.

Step 3: Select a role that defines the permissions of your lambda function. Select a new role with a basic lambda function and click on Create function.

Step 4: Inside the lambda function there is another option configuration. Go to configuration and click on permission. Then click on Role name.

Step 5: Attach AmazonTextractFullAccess and AWSLambdaExecute policies to the lambda permission role.

Step 6: Add S3 bucket as a trigger in lambda.

Step 7: Add code in lambda. Inside the code, we are using detect_document_text boto3 API which detects text in the input document. Amazon Textract API detects and analyses text in documents and converts it into machine-readable text. After adding the code save it and click on the deploy button. (GitHub Link)

Step 8: Upload one invoice image on the data-extract-from-image bucket.

Step 9: Check CloudWatch log groups. Inside the log event, you can get all your image extracted data.

Conclusion

In this blog, we learned about how to use AWS Textract API to extract data from an Image without any ML experience. This solution will drive decision-making efficiency and can be applied to any industry that has physical/scanned documents such as legal documents, purchase receipts, inventory reports, invoices, and purchase orders. We will discuss more use cases of AWS’s other services in our upcoming blogs.

About CloudThat