Skip to content
Trust Service Provider (eIDAS) · ISO 27001 · ISO 9001 +34 910 606 161
02 / 06  Computer vision & AI

Systems that see, check and decide in real time

We design and deploy computer vision and artificial intelligence systems for inspection, quality control, document reading and identity verification: our own models, integrated into your process and operated on our own infrastructure.

Flow of a computer vision system: capture, pre-processing, model, decision and integration OK 0.98 DEF 0.91 CAPTURE · INDUSTRIAL / MOBILE CAMERA Vision modeldetection · classification · OCR · biometrics Decisionbusiness rules · thresholds · traceability Process integrationPLC / MES / ERP · API · certified alerts · dashboard OPERATED 24/7 IN OUR OWN DATA CENTRE · ISO 27001 · DATA IN SPAIN
eIDAS Trust Service Provider
listed in the EU Trusted List
ISO/IEC 27001
information security, independently audited
ISO 9001
quality management, independently audited
Own infrastructure
high-performance data centres in Spain
Since 1998
28 years of uninterrupted engineering
Computer vision & AI

Applied computer vision: the model is half the job, integration is the other half

A computer vision system does not end with a model that scores well in the lab. It ends when the camera is well placed, the lighting is stable, the model runs within the time the line demands, the decision reaches the PLC or the ERP and every result is traced. That is the engineering we do.

We have spent years building and operating visual recognition systems in production: identity verification with document capture, OCR, MRZ reading, facial biometric matching and liveness detection, embedded in the remote contracting processes of thousands of companies; and visual inspection in industrial and logistics environments, where a false negative costs money.

We develop with our own and open-source models, trained on the client's data, and deploy them where they must run: at the edge (next to the camera), on your infrastructure or in our data centres, always with data processed in Spain under GDPR and our ISO 27001 certification.

Capabilities

  • Detection and classification of objects, defects and events in images and video.
  • OCR and document understanding: ID cards, passports, licences, invoices, delivery notes, number plates.
  • Facial biometrics with 1:1 matching and liveness detection against spoofing.
  • Measurement and counting of dimensions, parts, people or vehicles.
  • Edge or data-centre execution, with latencies matched to the line or the process.
  • Conversational AI agents on voice and chat, integrated with your data.
What is included

Computer vision and AI systems we build

From the camera to the business decision, with the traceability and security an audited process requires.

Inspection and quality control

Defect detection, assembly verification, presence/absence and in-line product classification, with a record of every decision.

Identity verification (KYC)

ID document + facial biometrics + liveness detection in under a minute, integrable via API into onboarding and remote contracting.

OCR and document reading

Data extraction from official documents, invoices and forms, with MRZ, format and consistency validation.

Real-time video recognition

Analysis of video streams for security, counting, number plates and events, with immediate alerts through Mensatek's channels.

Custom models

Training, evaluation and continuous improvement of models with the client's data, with agreed metrics and version control.

Industrial integration

Connection to PLC, MES, SCADA and ERP; synchronisation with the line; dashboards and traceability by batch.

Conversational AI agents

Assistants on chat, WhatsApp and voice that understand the user, query your data and hand over to a person when needed.

AI-based anti-fraud

Risk scoring of phones, emails, IPs and documents to detect impersonation and fraud in real time.

Operation and monitoring

Deployment at the edge or in our data centre, 24/7 monitoring of model quality and retraining when it drifts.

Technical guarantees

Accuracy measured, not promised

Every system is delivered with its real metrics on the client's data and with the means to watch them in production.

  • Agreed metrics: Precision, recall and false-positive rate defined before building and verified at acceptance.
  • Client data: Models trained and evaluated with real images from the process, not only with generic datasets.
  • Guaranteed latency: Edge or data-centre execution depending on the response time the line or the process demands.
  • Traceability: Every decision is logged with the image, the model, the version and the confidence: auditable afterwards.
  • Privacy and GDPR: Biometric and image data processed under GDPR and the ISO 27001 certification, in Spain.
  • Continuous improvement: Model drift monitoring and planned retraining, without stopping the process.
IDENTITY VERIFICATION · SESSION
Document captureFront and back · image quality
ok
OCR + MRZData read and mutually consistent
ok
1:1 facial matchingSelfie against the document photo
match
Liveness detectionPhoto, video or mask detection
real person
< 1 minTypical duration
APIIntegration
ESData in Spain
How we deliver it

From proof of concept to the system on the line

We start by measuring with your real images; we only build when the numbers hold.

Feasibility study

Analysis of the process, capture and lighting conditions, and proof of concept with real images.

System design

Camera and optics selection, execution architecture (edge or data centre), integration and target metrics.

Training and validation

Models trained and evaluated with the client's data; tests under production conditions.

Deployment and operation

Go-live, model quality monitoring and continuous improvement with the same team.

Typical use cases

Where it is applied

BANKING · INSURANCE · TELCO

Remote customer onboarding

Identity verification with document and biometrics embedded in the online contracting process, with certified evidence of the result.

INDUSTRY

In-line quality control

Visual inspection of parts and assemblies with a decision within cycle time and a record of every unit by batch.

LOGISTICS

Label and number-plate reading

OCR of delivery notes, labels and number plates to automate inbound, outbound and traceability.

PUBLIC SECTOR

Document processing

Extraction and validation of data from official documents to speed up case files with full traceability.

ENERGY · INFRASTRUCTURE

Installation inspection

Image analysis of installations and assets to detect anomalies and prioritise maintenance.

CUSTOMER SERVICE

AI agents

Conversational assistants on chat, WhatsApp and voice that resolve queries and hand over to people, with control over what the AI may and may not answer.

Frequently asked questions

Questions we are usually asked

Do we need many images to start?

Not for the feasibility study: with a representative set of real images we measure whether the problem is tractable and with what accuracy. For the final system the volume needed depends on process variability; we tell you with numbers after the proof of concept.

Where does the model run?

Wherever latency and security require: at the edge, next to the camera, when the decision must be taken in milliseconds; on your infrastructure; or in our data centres in Spain, with 24/7 monitoring.

How do you handle biometric data and images?

As specially protected data: processing under GDPR, minimisation, encryption, access control and audit logging, within our ISO 27001-certified management system, with data hosted in Spain.

What happens when the model is wrong?

It is designed for it: confidence thresholds, human review of doubtful cases and traceability of every decision. We watch the metrics in production and retrain when they degrade.

Does identity verification have legal validity?

The result of the process is documented in a certificate with qualified timestamping and can be incorporated into online contracting with electronic signature, within our eIDAS trust services.

Related areas

It works better combined

Do you have images and a process to improve?

Tell us what needs to be seen, under which conditions and what decision should be taken. We will propose a feasibility study with your own images.