Computer Vision (iBeta Level 3) • CV-01
Facial Recognition
1:1 verification and 1:N identification at gallery scale.
Scope
What the engagement covers.
Recognition accuracy in a laboratory rarely survives real cameras, real lighting and real populations. This capability is deployed with accuracy measured on the client’s own population and thresholds set against an explicit error trade-off.
Included capabilities
- 1:1 verification for authentication and document matching
- 1:N identification with sub-second search across large galleries
- Template versioning and re-enrolment strategy as models improve
- Threshold policy tuned to the accepted false accept and reject balance
- Tolerance for pose, ageing, eyewear and partial occlusion
Outputs and deliverables
- Accuracy benchmark on client population
- Gallery and template architecture
- Threshold policy with error trade-off rationale
- Ongoing accuracy and bias monitoring reports
Workflow
How it is delivered, step by step.
Each step has an owner, an entry condition and an artefact that has to exist before the next step begins.
01DefineUse case, population, accuracy targets and error tolerance.
02BenchmarkAccuracy testing on representative client data.
03DesignGallery structure, template storage and threshold policy.
04DeployIntegration, tuning and human adjudication workflow.
05MonitorAccuracy and demographic performance tracked over time.
Use cases
Where this is typically applied.
Customer authentication in branch or at self-service points
De-duplication of an existing enrolment registry
Delivery model
The operating pattern for Computer Vision (iBeta Level 3).
The same delivery discipline applies across every capability in this line, so combined engagements stay coherent.
Define
Use case, lawful basis, accuracy targets and acceptance thresholds.
Prove
Bench test on representative data, including attack and edge-case sets.
Integrate
SDK or API embedding, gallery design and decision policy.
Deploy
Edge or on-premise rollout, tuning and human adjudication workflow.
Assure
Accuracy monitoring, bias review, revalidation and retention enforcement.
Integration
- Core banking, onboarding and case management platforms via SDK or API.
- Physical access controllers, turnstiles and video management systems.
- Identity registries and watchlists under a controlled matching policy.
- SIEM and audit platforms for decision evidence and administrative logs.
Engagement approach
Every deployment is scoped with a documented lawful basis, retention schedule and human adjudication path before a single template is enrolled. Accuracy and bias are measured on the client’s own population, not on vendor benchmarks.
Related
Other capabilities in Computer Vision (iBeta Level 3).
CV-02
Liveness & Anti-Spoofing
Prove there is a live person in front of the camera.
CV-03Digital Identity
A reusable, governed identity credential instead of repeated verification.
CV-04KYC Onboarding
Open the account remotely without opening a fraud channel.
CV-05Face, Body & Object Recognition
Understand what is in the scene beyond identity.
CV-06Vehicle Analytics
Read, classify and track vehicles automatically.
CV-07Smart City
Operational intelligence for public space at city scale.
CV-08Video Forensics
Find the thirty seconds that matter in ten thousand hours.
CV-09Border & Public Safety
High-assurance identity at controlled crossing points.
CV-10SDK / API Integration
Put biometric capability inside the applications people already use.