Step 1
1. You upload a scan or a photo
A document from a scanner, an MFP or a phone camera — one by one or in a batch through the API.
Handwriting recognition in forms and documents
NeuroVision turns handwritten documents into data: names, dates, addresses and amounts from forms, applications and blanks filled in by hand arrive in your system as ready JSON fields.
Recognition result Handwritten form
- Last name high IVANOV
- First name high IVAN
- Date of birth high 12.03.1985
- Phone high +7 912 345-67-89
- Address review Moscow, Lesnaya St., 7
- Fields with a confidence score
- REST API, SDK and on-premises deployment
- A free month on your own forms
Data is stored in Russia Russian Software Registry No. 19972 Compliance with 152-FZ
What we recognise in handwritten documents
Fields filled in block letters — in boxes or on the lines of a form — are recognised best. The service returns a confidence score for every field, and doubtful fields can be sent to an operator.
-
Forms and applications
Customer forms, loan and payout applications, consents. We extract names, dates, addresses, phone numbers and checkbox marks.
-
Forms with figures and amounts
Handwritten figures, dates and amounts in receipts, waybills and acts. The value format is checked: a date as a date, a tax number against its check digit.
-
Handwritten entries in a passport
Pages with registration and stamps where part of the entries is handwritten. Passport recognition — a separate page.
-
Medical documents
Patient forms, informed consents, referrals: the data reaches the clinic's system without retyping.
-
Reports and acceptance acts
Road accident reports, handover acts, waybills — documents that are often filled in by hand on the spot.
-
Signatures and marks
We check that the signature and the required marks are in place. The signature is not «read» — its presence is recorded.
How handwriting recognition works
The neural network reads the whole field and checks the result against the format and reference data. A person steps in only for fields the model is unsure about.
Step 2
2. We locate fields by the form template
A template is set up for your forms: where the name is, where the date is, where the amount is. The model aligns the image and finds every field.
Step 3
3. We recognise the handwriting
The neural network reads characters taking neighbours and the meaning of the field into account: a «Date» field expects figures, a «Last name» field expects letters.
Step 4
4. We check the format and reference data
Dates, phone numbers and document numbers are checked against their format, names and addresses against reference data.
Step 5
5. We return the result in JSON
A value and a confidence score for every field. Fields with low confidence can be routed to an operator automatically.
OCR, ICR and cursive handwriting: what recognises what
OCR (optical character recognition) reads printed text. Handwriting needs ICR (intelligent character recognition) — recognition of handwritten characters written separately. Cursive is the hardest to recognise, which is why forms meant for automatic processing ask to be filled in block letters.
| Criterion | Printed text (OCR) | Block letters (ICR) | Cursive handwriting |
|---|---|---|---|
| Example | passport, contract, statement | a form with boxes, an application blank | a letter, a note, lecture notes |
| How it is recognised | character by character, with high accuracy | character by character, using the meaning of the field | whole words, accuracy depends on the handwriting |
| When a person is needed | rarely | for fields with low confidence | often |
| NeuroVision | yes, document recognition | yes | on request, on your own samples |
Handwriting recognition API
Batch processing of scans through the REST API, or capturing a form in an app through the SDK. For archives with personal data — deployment inside your own perimeter.
- Fields in JSON
- Confidence for every field
- Templates for your forms
- Batch processing
Terms of connection
A free one-month test on your own forms, a dedicated manager, 24/7 support, payment only for processed documents.
Where handwriting recognition is used
Everywhere customers and employees still fill in paper and operators retype it into the system.
-
Banks and MFIs
- Applications and forms filled in a branch or at an agent
- Data in the core banking system without retyping
-
Insurance
- Road accident reports and claim applications
- Faster claims settlement
-
Healthcare
- Patient forms and informed consents
- Data in the clinic's system and shorter queues at the front desk
-
Government and service centres
- Citizens' applications filled in by hand
- Processing without manual entry
-
Logistics
- Waybills, acts and route sheets with handwritten entries
- Data in the accounting system on the day of delivery
-
Archives and HR departments
- Digitising personnel files and paper archives
- Search by data, not by folders
Why a business needs automatic handwriting recognition
-
Minutes instead of hours
An operator spends several minutes retyping a handwritten form. The service processes a batch of scans while the operator checks only the doubtful fields.
-
Fewer errors in the data
Every field is checked against its format, and uncertain values do not reach the system unchecked.
-
The archive becomes data
Paper forms and applications can be searched, filtered and exported into reports.
Operators are relieved. Automation reads the standard fields,
people are left only with the illegible ones.
How much handwritten document recognition costs
The price depends on the volume and the number of forms. The first month is a free test on your own forms.
Calculate the cost- 01
Payment per processed page, no subscription fee
- 02
Template setup for your forms
- 03
Cloud or deployment inside your perimeter
Why NeuroVision
Handwriting recognition is part of the NeuroVision platform: the same models, infrastructure and security requirements.
NeuroVision — an expert in computer vision and AI
AI-OCR works as part of customer verification (KYC), trusted by the largest companies in fintech, e-commerce and transport services.
We develop identification algorithms that automate checks, detect fraud and help meet regulatory requirements.
A Russian developer
A Skolkovo resident, software in the Russian Software Registry (No. 19972), Rospatent patent No. 2023661508.
Licences and certificates
The project «Development of a multimodal system using artificial intelligence "Neuromed" for managing the diagnostic processes of a clinic on the basis of the "Neurovision" computer vision platform» is supported by the Innovation Promotion Foundation within the federal project «Domestic Solutions»
The Neurovision computer vision platform is registered with Rospatent (certificate No. 2023661508) and included in the Russian Software Registry (entry No. 19972 of 13.11.2023)
The right to use the Platform is granted under a licence agreement concluded by Neurovision with its customers
Test recognition on your own forms
Send us a few filled-in forms without personal data, or form samples — we will show how the service recognises them and give you a month of test access.
Can't put your request into words? Book a call and our team will contact you at a convenient time
Book a call
The company is included in the register of IT companies and the register of personal data operators of the Russian Ministry of Digital Development.
Address: Moscow, Presnenskaya Embankment 12 — Moscow City business centre, Federation Tower. Nearest metro stations: Delovoy Tsentr and Mezhdunarodnaya.
Read also
Materials on document recognition and AI-OCR accuracy — in our blog.
331 23
KYC Document Verification
AI-OCR in KYC: how automatic document recognition reduces manual review
377 10
AI OCR, IDP and Document Recognition
How to improve the accuracy of AI OCR and MRZ verification in KYC: metrics, errors and working solutions
352 14
KYC, AML and Digital Identification
How AI OCR is changing KYC: the technology, accuracy and role in digital verification
363 21
KYC Document Verification
Why users retake documents: how to reduce the number of repeat attempts during a KYC check
Frequent questions about handwriting recognition
Answers about handwriting, accuracy, document types, data storage and pricing
OCR recognises printed text, ICR recognises handwritten characters. ICR takes into account that the same letter looks different from person to person, and checks the result against the meaning of the field.
Fields filled in block letters are recognised most reliably. For cursive we recommend testing the service on your own samples: accuracy depends on the handwriting and the quality of the scan.
Accuracy depends on the form, the handwriting and the quality of the scan. The service returns a confidence score for every field, and uncertain values can be sent to an operator.
Forms, applications and blanks filled in by hand, as well as handwritten entries in a passport. A template is set up for your forms, and the service finds the required fields itself.
NeuroVision is a service for companies, not an online tool for individuals. For companies we provide a free one-month test on their own forms.
Fields with a box for each letter, a request to fill them in block letters with a dark pen, and a scan or photo without glare or shadows.
Data is processed on servers in Russia in accordance with 152-FZ. With deployment inside your perimeter, documents never leave your infrastructure.
Payment is per processed page, and the first month is a free test. We will calculate the exact cost for your volume and forms.
