2. Creating an account to access Text Recognition Software on Cloud Platforms
Google’s GCP Cloud Vision and Microsoft’s Azure Cognitive Computer Vision API are (as of Fall 2020) the two best cloud OCR-HTR services for recognizing and extracting a wide array of printed and handwritten text. Handprint is capable of using Amazon’s Rekognition and Textract services but since I have yet to encounter a use case for which Amazon’s services have proven more accurate, this tutorial will focus on the use of GCP Cloud Vision and Azure Computer Vision. Both GCP and Azure offer free usage accounts with 1000 and 5000 free units (respectively) for OCR-HTR use per month (think 1000 and 5000 pages/images per month free of charge). They also offer free credits for new users and grants to students, faculty, and staff for additional credits.
2a. Set Up Microsoft Azure Cloud for Text Recognition and Extraction
To sign up for Azure and obtain a key, visit https://portal.azure.com and create a new account. (Note: you will need to turn off browser security plugins such as Ad Block and uMatrix if you have them, or else the site will not work.) You can use a version of your Brown Credentials to login to Azure by entering your MyAccount username plus @ad.brown.edu (ex. jcarberry@ad.brown.edu). You will then be redirected to the Microsoft Azure Dashboard https://portal.azure.com, from where you can create credentials. Otherwise, if you create a new account, after going through the various screens, you will also end up on the Dashboard page.
Before you create any service accounts, the Microsoft Azure Dashboard looks somewhat like this:
To create a service account, first click on the green Create Resources button. This will lead to a page where you select resources that looks somewhat like this:
Click AI + Machine Learning; this will bring up a list of “Featured” options in the right-hand column (the things with the blue squares in the screen capture above), and there look for Computer Vision. Click on that one.
The next page will look somewhat like this:
Click Start Free, then click through to fill out account information and identification. It will require a phone number and a credit card, but you won’t be charged anything.
After all that’s done, you will once again be put back to the Portal Dashboard. Click on Create Resources again, select AI + Machine Learning followed by Computer Vision again, and this time the following form will appear:
Fill in the blanks in this form. Give the service a name (for your own organizational purposes), keep the “Free Trial”, and for the location, make sure to select West US. For the pricing tier, select the F0 free option. For the “Resource group”, click on the “Create new” link below the blank and create a new resource name – it can be anything that makes sense, such as “Handprint” or “OCR test” or whatever.
Finally, after the form is done, you should be able to click on the resource name in your Azure Dashboard. This should lead to a page that looks somewhat like this:
Double-check what the API endpoint begins with (this should be https://westus.api.cognitive.microsoft.com) in Section 2 and then click on the Keys link in Section 1. The key will be a string such as "18de248475134eb49ae4a4e94b93461c". The next page will provide two keys. You can copy either one and save it in a file myazurecredentials.json.
➩Continue to Step 2b. Set up Google Cloud Platform for text recognition and extraction