DocumentAI FormParser fails with UnknownError for forms

[Bedlam]

For forms of certain length (originally excel tables) provided as PNG images, formParser fails with 'unknown error'.

How to reproduce:

• Go to https://cloud.google.com/document-ai/docs/drag-and-drop
• Select FormParser
• add any of the attached file(s) (see bottom of post)

Response:

The file does not meet any specs mentioned in DocumentAI/Form parser limitations

---

Test files:

[Poala_Tenorio]

The "Unknown Error" response from Google Cloud's DocumentAI FormParser can be challenging to diagnose without specific error details. However, if you are consistently encountering this error with certain images, especially when they were originally Excel tables, it might be related to the format or quality of the images being processed.

Make sure that your API request is correctly formatted and includes all the required parameters. This includes specifying the input document, enabling FormParser in your request, and providing any additional options or settings as needed.

If you are dealing with Excel tables, ensure that any tables or tabular structures are properly detected within the image. The FormParser relies on table structure to extract data, so if the tables are not detected, it can result in an error. You can try highlighting or outlining the tables in the image to improve detection.

Ensure that the input document matches the capabilities of the FormParser. FormParser is designed to extract structured data from forms, so the document should be in a format that resembles a form. If the document is too unstructured or doesn't contain form-like elements, it may not work effectively.

The quality and legibility of the document can impact the parsing results. Make sure the document is clear and well-scanned. If it's a handwritten form, it may be more challenging to parse accurately.

Check and make sure you haven't exceeded your API quotas or usage limits. You can check your API usage and quotas in the Google Cloud Console.

[Bedlam]

Hi Poala,

Thank you for responding. In this scenario, the exact same image worked before September 8 (it was used in a regression test scenario) and the behavior changed afterwards.

For invalid argument, it's not clear what the requirements are. We are aware of the max size limits from DocumentAI and FormParser, but not any specifics wrt the form in the image. Quota's/usage limits have been checked and the API works for other images.

Kind regards,