Layout Parser in Document AI: Boundin box issue

[vguleria]

I'm using the Layout Parser in Document AI, but it doesn't seem to return bounding box details. Is there a way to retrieve these details from the API?

Currently, the response I receive looks like this:

{
  "document": {
    "documentLayout": {
      "blocks": [
        {
          "blockId": "1",
          "textBlock": {
            "text": "Firefox",
            "type": "header"
          },
          "pageSpan": {
            "pageStart": 1,
            "pageEnd": 1
          }
        }
      ]
    }
  }
}

[ruthseki]

Hi @vguleria,

Welcome to Google Cloud Community!

Here's what you should look for:

• boundingPoly: This is the most common field name for the bounding box. It usually contains normalized coordinates relative to the page. 
• layout (Optional): In some older versions or slightly different API implementations, bounding box information might be nested within a layout object.

You may try these approaches to fix your code:

1. Check your API response carefully: Examine the complete JSON response from the Document AI Layout Parser API. Search for "boundingPoly" or "layout" within the blocks array. The boundingPoly is usually a direct sibling of blockId, textBlock, and pageSpan.
2. Update your code to access the bounding box: Once you've located the boundingPoly field, update your code to extract the vertex coordinates. 

Additionally, you may check these links:

• How to use Document AI
• Handle processing response
• Process documents with Layout Parser
• Package google.cloud.documentai.v1
• Package google.cloud.documentai.v1beta3

I also found this case that you might find useful.

I hope the above information is helpful.

[nguyen_toan]

Hello friend, I also have the same problem as you. You have now solved the above problem. Can you please tell me how to fix them if you have solved them.

[divekarsc]

Has someone found solution to this original question?

[verasupport]

It appears that you `returnBoundingBoxes` is exposed in `layoutConfig` but I'm still not seeing it in the return payload.  `returnImages` is working as expected. 

 const request: protos.google.cloud.documentai.v1beta3.IProcessRequest = {
    name,
    rawDocument: {
      content: documentContent,
      mimeType,
    },
    processOptions: {
      layoutConfig: {
        returnBoundingBoxes: true, // not working
        returnImages: true,
      },
    },
  };

  const [result] = await client.processDocument(request);

@ruthsekiAny ideas?