Skip to content

BETASIGNATURE DETECTION

Find signature candidates
in scanned documents.

Detect handwritten signature regions before a document moves to the next step. Use ChequeDB's beta to find candidate signatures across scanned forms, agreements, mandates and cheques, then connect their positions to the checks your workflow requires.

Forms · contracts · mandates · delivery notes · cheques

Document → signature evidenceILLUSTRATION
Synthetic agreement: two signature drawingsA fictional agreement, not a customer document. Candidate one is at x 36, y 378, width 154, height 75. Candidate two is at x 230, y 378, width 154, height 75. No inference is performed.SYNTHETIC DOCUMENT01 / 01Service agreementExample document for a beta evaluationSCOPE & ACCEPTANCECANDIDATE 01CANDIDATE 02Applicant fieldCountersigner field

THE FOUR QUESTIONS

Two candidate regions

Two drawn signature candidates appear on this fictional agreement. Their positions do not identify the signers.

Synthetic drawings and example workflow. No upload, model inference or live validation is performed here.

START WITH PRESENCE

Check for a signature before comparing one.

Detection can be useful even when your organisation holds no signature specimens. It identifies candidate regions in the page image; positioning shows where they occur. That gives your process a starting point for checking a required field or presenting the relevant area to a reviewer.

Reference samples become necessary when the question changes to whether the candidate resembles the writing held for a particular signer. Keep that verification step separate so a presence-only workflow does not depend on reference collection.

SYNTHETIC EXAMPLES

One signature field can need three different actions.

Imagine a fictional consent form with a signature line near the foot of the page. These situations explain detection and handling; the drawings do not run live inference.

A document with no detected candidate should not automatically be labelled a mismatch. There may be nothing to compare, or the image may not contain enough usable evidence.

Example imageWhat the evidence supportsWhat to resolve next
Visible handwritten markA candidate can be inspected in its page context.Is it in the required field, and is comparison needed?
Apparently blank fieldNo visible signature appears in the illustration.Check that the page is complete and readable before treating it as unsigned.
Stamp across the writingA possible signature is partly obscured.Inspect the original page or obtain a clearer image.

DOCUMENT CONTEXT

Count candidates and required signatures separately.

A form might require an applicant and a guardian signature. Two candidate regions on the page do not establish that both required people signed. One could be a note, an initial or a mark in another field.

In the beta evaluation, identify the required areas and how your workflow associates each candidate with a field. Include optional signature blocks, amended layouts and several nearby marks. Where assignment is ambiguous, agree who inspects the page before it proceeds.

EVALUATE THE DIFFICULT CASES

Test the marks your documents actually contain.

01

Stamps and printed backgrounds

Include stamps crossing signature lines and dense text around the field. Finding a mark and extracting usable writing are separate questions.

02

Initials and handwritten notes

Label which marks count as signatures for the document task. A short initial and a full signature may require different handling.

03

Faint or incomplete scans

Include washed-out ink, clipped page edges and soft images. Capture problems can explain an absent or uncertain candidate.

04

Unexpected placement

Include signatures above the line, in a margin or on a different page. Decide how location affects the document check.

A PRACTICAL EVALUATION

Label the page before measuring the detector.

Create a permissioned sample set from representative layouts. Include documents with signatures and documents without them, then record what a useful detection should find.

Measure how many expected signatures are found and how many detected regions are incorrect. Review the full pages behind those errors; a single headline accuracy number can hide the cases that create manual work.

RecordWhy it matters
Expected signature regions on each pageShows whether candidates were missed or placed incorrectly.
Marks that should not countShows whether text, stamps or notes produce extra detections.
Required versus optional fieldsSeparates detection performance from the document's completion rule.
Capture quality and layoutMakes failures visible in the conditions your operation encounters.

QUESTIONS & ANSWERS

Detection: practical questions.

Can detection work without a reference signature?

Yes. Detection identifies candidate handwritten-signature regions in the page image. Authorised reference samples are needed for verification against a claimed signer, not for finding a candidate.

Does a detected signature tell us who signed?

No. A detected region describes a candidate's presence and location. Associating it with a field or signer requires the document context and your reference or operational process.

Can we check documents with more than one signature?

Include those layouts in the beta scope. Agree the required fields, how candidates are assigned to them and what happens when the count or placement is ambiguous. Candidate count alone does not establish the required signers.

What if no signature is detected?

Distinguish an apparently blank field from a poor or incomplete scan and a missed candidate. The next action should follow your document requirements and review policy.

CHEQUEDB SIGNATURES · BETA

Start with your documents.
Define the checks that matter.

Discuss your document types, signature references and validation workflow with the ChequeDB team.