CheckID
CustomersPartners
VIEN
Contact
BIOMETRIC SCIENCE

Face — from image to evidence-backed decision

Face is more than a selfie. exposes capture, quality, representation, matching, liveness/PAD, threshold and evidence, while separating standards concepts from Mobile-ID measured performance.

Talk to an expertTechnical documentation
◈
SOURCEAuthoritative / document
◈
CAPTUREDevice / SDK
◈
VERIFYDocument + biometric
◈
ASSUREPolicy + risk
◈
DECIDEClaims / status
◈
EVIDENCEAudit / trace
TECHNICAL REFERENCE

Face: capture envelope, holder binding and matching

Portrait/live capture → quality → PAD → template/match → policy/evidence

Image source

Face journeys must distinguish a portrait obtained from a document/chip from a live image or video. These sources have different quality, purpose and threat models.

Capture quality

Relevant dimensions include face detection, usable face size, pose, illumination, blur, occlusion and multiple faces. Current documentation publishes no pose angle or minimum pixel requirement without a tested profile.

PAD placement

PAD addresses presentation attacks on the capture path within its implementation/evaluation scope. A PAD outcome is not a match score and does not replace 1:1 holder binding.

1:1 verification

Holder binding compares the live representation with a reference portrait/template. Threshold profiles and calibration must be tied to deployment/test context; the ZIP does not support a numeric threshold.

1:N identification

Where `FACIAL_IDENTIFICATION` is used, 1:N is kept distinct from 1:1. Gallery construction, candidate behavior and adjudication need their own evidence. The API catalog supports the process/group concept but does not provide benchmark data.

Failure modes

No face, multiple faces, insufficient quality, PAD inconclusive/fail, missing reference, match below policy threshold and unavailable capture channel are distinct failure classes to be mapped to actual errors/results when a source exists.

Authoritative technical references