Can the Custom Neural Model Extract Signatures?

Question

In the documentation, it is mentioned that the custom template model can extract signatures, while the custom neural model is stated to only detect signatures.

The relevant documentation links are:

• Custom Template Documentation "Custom template (formerly custom form) is an easy-to-train document model that accurately extracts labeled key-value pairs, selection marks, tables, regions, and signatures from documents."
• Custom Neural Model Documentation "Custom neural v4.0 2024-11-30 (GA) model supports signature detection, table cell confidence, and overlapping fields."

Is it possible for the custom neural model to extract signatures as well?

Answer 1

Hi Tyrone Wilkinson

The Custom Neural Model v4.0 GA does support signature fields. To ensure that signature detection works correctly. Here are some steps to follow.

• Use the Signature field type and accurately draw the region for each signature, ensuring only one draw region per field.
• Train the model with at least five labeled samples of signatures, including variations in signature styles to improve accuracy.
• If your trained model isn't detecting signatures, check that the signatures were correctly labeled in your training data.
• Review the model’s raw JSON output to see if signature fields are present but have low confidence scores.
• Use document samples that closely match the format of your training data during testing.

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/train/custom-neural?view=doc-intel-4.0.0&form=MG0AV3

Hope this helps. Do let us know if you any further queries.

---

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.