Edit

Get started: Form Recognizer Studio

  • Article

This article applies to: Form Recognizer v3.0 checkmark Form Recognizer v3.0. Earlier version: Form Recognizer v2.1

Form Recognizer Studio is an online tool for visually exploring, understanding, and integrating features from the Form Recognizer service in your applications. You can get started by exploring the pretrained models with sample or your own documents. You can also create projects to build custom template models and reference the models in your applications using the Python SDK and other quickstarts.

Prerequisites for new users

Tip

Create a Cognitive Services resource if you plan to access multiple cognitive services under a single endpoint/key. For Form Recognizer access only, create a Form Recognizer resource. Please note that you'll need a single-service resource if you intend to use Azure Active Directory authentication.

Models

Prebuilt models help you add Form Recognizer features to your apps without having to build, train, and publish your own models. You can choose from several prebuilt models, each of which has its own set of supported data fields. The choice of model to use for the analyze operation depends on the type of document to be analyzed. Form Recognizer currently supports the following prebuilt models:

Document analysis

  • General document: extract text, tables, structure, key-value pairs.
  • Layout: extract text, tables, selection marks, and structure information from documents (PDF, TIFF) and images (JPG, PNG, BMP).
  • Read: extract text lines, words, their locations, detected languages, and handwritten style if detected from documents (PDF, TIFF) and images (JPG, PNG, BMP).

Prebuilt

  • Invoice: extract text, selection marks, tables, key-value pairs, and key information from invoices.
  • Receipt: extract text and key information from receipts.
  • Health insurance card: extract insurer, member, prescription, group number and other key information from US health insurance cards.
  • W-2: extract text and key information from W-2 tax forms.
  • ID document: extract text and key information from driver licenses and international passports.
  • Business card: extract text and key information from business cards.

Custom

  • Custom extraction models: extract information from forms and documents with custom extraction models. Quickly train a model by labeling as few as five sample documents.
  • Custom classification model: train a custom classifier to distinguish between the different document types within your applications. Quickly train a model with as few as two classes and five samples per class.

Gated preview models

Note

To request access for gated preview models in Form Recognizer Studio, complete and submit the Form Recognizer private preview request form.

  • General document with query fields: extract labels, values such as names, dates, and amounts from documents.
  • Contract: extract the title and signatory party information (including names, references, and addresses) from contracts.
  • Vaccination card: extract card holder name, health provider, and vaccination records from US COVID-19 vaccination cards.
  • US 1098 tax form: extract mortgage interest information from US 1098 tax forms.
  • US 1098-E tax form: extract student loan information from US 1098-E tax forms.
  • US 1098-T tax form: extract tuition information from US 1098-T forms.

Note

To request access for gated preview models in Form Recognizer Studio, complete and submit the Form Recognizer private preview request form.

After you've completed the prerequisites, navigate to Form Recognizer Studio General Documents.

In the following example, we use the General Documents feature. The steps to use other pretrained features like W2 tax form, Read, Layout, Invoice, Receipt, Business card, and ID documents models are similar.

Selecting the General Document API to analysis a document in the Form Recognizer Studio.

  1. Select a Form Recognizer service feature from the Studio home page.

  2. This step is a one-time process unless you've already selected the service resource from prior use. Select your Azure subscription, resource group, and resource. (You can change the resources anytime in "Settings" in the top menu.) Review and confirm your selections.

  3. Select the Analyze button to run analysis on the sample document or try your document by using the Add command.

  4. Use the controls at the bottom of the screen to zoom in and out and rotate the document view.

  5. Observe the highlighted extracted content in the document view. Hover your mouse over the keys and values to see details.

  6. In the output section's Result tab, browse the JSON output to understand the service response format.

  7. In the Code tab, browse the sample code for integration. Copy and download to get started.

Added prerequisites for custom projects

In addition to the Azure account and a Form Recognizer or Cognitive Services resource, you need:

Azure Blob Storage container

A standard performance Azure Blob Storage account. You create containers to store and organize your training documents within your storage account. If you don't know how to create an Azure storage account with a container, following these quickstarts:

  • Create a storage account. When creating your storage account, make sure to select Standard performance in the Instance details → Performance field.
  • Create a container. When creating your container, set the Public access level field to Container (anonymous read access for containers and blobs) in the New Container window.

Configure CORS

CORS (Cross Origin Resource Sharing) needs to be configured on your Azure storage account for it to be accessible from the Form Recognizer Studio. To configure CORS in the Azure portal, you need access to the CORS tab of your storage account.

  1. Select the CORS tab for the storage account.

    Screenshot of the CORS setting menu in the Azure portal.

  2. Start by creating a new CORS entry in the Blob service.

  3. Set the Allowed origins to https://formrecognizer.appliedai.azure.com.

    Screenshot that shows CORS configuration for a storage account.

    Tip

    You can use the wildcard character '*' rather than a specified domain to allow all origin domains to make requests via CORS.

  4. Select all the available 8 options for Allowed methods.

  5. Approve all Allowed headers and Exposed headers by entering an * in each field.

  6. Set the Max Age to 120 seconds or any acceptable value.

  7. Select the save button at the top of the page to save the changes.

CORS should now be configured to use the storage account from Form Recognizer Studio.

Sample documents set

  1. Go to the Azure portal and navigate as follows: Your storage account → Data storage → Containers

    Screenshot: Data storage menu in the Azure portal.

  2. Select a container from the list.

  3. Select Upload from the menu at the top of the page.

    Screenshot: container upload button in the Azure portal.

  4. The Upload blob window appears.

  5. Select your file(s) to upload.

    Screenshot: upload blob window in the Azure portal.

Note

By default, the Studio will use form documents that are located at the root of your container. However, you can use data organized in folders by specifying the folder path in the Custom form project creation steps. See Organize your data in subfolders

Custom models

To create custom models, you start with configuring your project:

  1. From the Studio home, select the Custom model card to open the Custom models page.

  2. Use the "Create a project" command to start the new project configuration wizard.

  3. Enter project details, select the Azure subscription and resource, and the Azure Blob storage container that contains your data.

  4. Review and submit your settings to create the project.

  5. From the labeling view, define the labels and their types that you're interested in extracting.

  6. Select the text in the document and select the label from the drop-down list or the labels pane.

  7. Label four more documents to get at least five documents labeled.

  8. Select the Train command and enter model name, select whether you want the custom template (form) or custom neural (document) model to start training your custom model.

  9. Once the model is ready, use the Test command to validate it with your test documents and observe the results.

Form Recognizer Custom model demo

Labeling as tables

Note

  • With the release of API versions 2022-06-30-preview and later, custom template models will add support for cross page tabular fields (tables).
  • With the release of API versions 2022-06-30-preview and later, custom neural models will support tabular fields (tables) and models trained with API version 2022-08-31, or later will accept tabular field labels.

  1. Use the Delete command to delete models that aren't required.

  2. Download model details for offline viewing.

  3. Select multiple models and compose them into a new model to be used in your applications.

Using tables as the visual pattern:

For custom form models, while creating your custom models, you may need to extract data collections from your documents. Data collections may appear in a couple of formats. Using tables as the visual pattern:

  • Dynamic or variable count of values (rows) for a given set of fields (columns)

  • Specific collection of values for a given set of fields (columns and/or rows)

Label as dynamic table

Use dynamic tables to extract variable count of values (rows) for a given set of fields (columns):

  1. Add a new "Table" type label, select "Dynamic table" type, and name your label.

  2. Add the number of columns (fields) and rows (for data) that you need.

  3. Select the text in your page and then choose the cell to assign to the text. Repeat for all rows and columns in all pages in all documents.

Form Recognizer labeling as dynamic table example

Label as fixed table

Use fixed tables to extract specific collection of values for a given set of fields (columns and/or rows):

  1. Create a new "Table" type label, select "Fixed table" type, and name it.

  2. Add the number of columns and rows that you need corresponding to the two sets of fields.

  3. Select the text in your page and then choose the cell to assign it to the text. Repeat for other documents.

Form Recognizer Labeling as fixed table example

Signature detection

Note

Signature fields are currently only supported for custom template models. When training a custom neural model, labeled signature fields are ignored.

To label for signature detection: (Custom form only)

  1. In the labeling view, create a new "Signature" type label and name it.

  2. Use the Region command to create a rectangular region at the expected location of the signature.

  3. Select the drawn region and choose the Signature type label to assign it to your drawn region. Repeat for other documents.

Form Recognizer labeling for signature detection example

Next steps

Get started with the Form Recognizer Studio.