Share via

Using My Own Labeled Data with Azure Document Intelligence Custom Neural Model

Pedro Preto 0 Reputation points
2024-06-03T16:18:54.5933333+00:00

Hi,

I want to use my own labeled data with Azure Document Intelligence Custom Neural Model.

I have labeled a series of receipt images using my own labeling tool. Can I use this labeled data to train a custom neural model? If yes, could you provide details on the required file formats (fields.json, {file}.ocr.json, {file}.labels.json) so that I can convert my data accordingly?

Thanks in advance!
Pedro

Azure AI Document Intelligence
Azure AI Document Intelligence
An Azure service that turns documents into usable data. Previously known as Azure Form Recognizer.
1,475 questions

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. VasaviLankipalle-MSFT 15,401 Reputation points
    2024-06-03T23:05:03.3066667+00:00

    Hello @Pedro Preto , Thanks for using Microsoft Q&A Platform.

    Yes, you can use your own labeled data to train a custom neural model with Azure Document Intelligence. We strongly recommend using DI Studio for label creation due to its proven quality and reliability. When you manually update labels, there's a significantly higher risk of introducing errors and inconsistencies.

    DI Studio automates this process, ensuring accuracy and minimizing the likelihood of issues that can arise from manual updates. This streamlines the workflow, saving time and reducing the potential for mistakes.

    To do so, you will need to upload your labeled data to an Azure blob storage container. The required file formats for training a custom neural model with Azure Document Intelligence are:

    • A fields.json file is created when the first field is added. There's one fields.json file for the entire training dataset, the field list contains the field name and associated sub fields and types.
    • The Studio runs each of the documents through the Layout API. The layout response for each of the sample files in the dataset is added as {file}.ocr.json. The layout response is used to generate the field labels when a specific span of text is labeled.
    • A {file}.labels.json file is created or updated when a field is labeled in a document. The label file contains the spans of text and associated polygons from the layout output for each span of text the user adds as a value for a specific field.

    I hope this helps.

    Regards,

    Vasavi

    -Please kindly accept the answer and vote 'yes' if you feel helpful to support the community, thanks.