AWS Machine Learning: Building an Expense Tracker Using Amazon Textract

24mIntermediate2021-02-10

Authors

Carlos Rivas

AWS Infrastructure Expert

Course details

If you’ve ever used an optical character recognition (OCR) program, you know that scanning documents for text can often be hit or miss. If you’re a developer working on an app that's dependent on accurate text scanning, say for an expense report, hit or miss just won’t cut it. In this project-based course, Carlos Rivera shows you how to use Amazon Textract to analyze scanned documents and convert them to text. Textract eliminates the complexity of having to train machine learning models from scratch to perform data capture tasks. And as Carlos points out, the program not only recognizes text, but it also considers the layout of the scanned document. Follow along with Carlos as he creates a serverless expense tracker that reads text from images using Textract, starting with the basic jargon of the program, through project implementation, and then implementing Textract.

Skills covered

Machine LearningAmazon Web Services (AWS)AmazonPersonaCloud ServicesCloud PlatformsArtificial Intelligence (AI)Cloud Computing

Concepts

0. Introduction

01 - Machine learning for optical character recognition

1. Architecture

02 - Textract concepts
03 - AWS Textract overview
04 - Expense tracker architecture

2. Project Implementation

05 - Implementing an AWS blueprint to integrate Lambda with S3
06 - Using S3 uploads to trigger a Lambda function
07 - Integrating Textract into the Python Lambda
08 - Using Textract in Python to process an image

3. Textract Implementation

09 - Parsing Textract metadata to get the required information
10 - Using regular expressions to find the desired values
11 - Looking for keywords within the extracted text

4. Totaling

12 - Updating JSON file with the current receipt total
13 - Using S3 as persistent storage for receipt details
14 - Validating and summarizing several executions of the code

Conclusion

15 - Next steps