Biometrics Available
Atlas Face
Match the selfie to the document photo, with an explainable match score.
Access by invitation: no account yet? Ask for your sandbox workspace.
What it checks
The face on the selfie, matched to the one on the document.
Atlas Face finds the portrait on the document, compares it with the selfie and tells you why it concludes what it does.
-
Face detection
The portrait is located on the document and on the selfie before any comparison.
-
Match score
One score per verification, compared with your threshold; below it, the file goes to review.
-
Adjustable threshold
You set the threshold to match your risk appetite, and can change it at any time.
-
Several faces
A selfie with several people is flagged rather than compared at random.
-
Reused portrait
The document photo presented as a selfie is recognised and rejected.
-
Biometrics deleted
Selfie and face template deleted when the file closes, unless you choose otherwise.
How it works
One comparison, one score, one explanation.
- 1 Document read Atlas ID extracts the portrait printed on the document.
- 2 Guided selfie The customer centres their face in the oval; quality is checked.
- 3 Comparison The two faces are compared and a score is computed.
- 4 Decision Above the threshold, the check passes; below it, an analyst decides.
Face matchSample data
Document photo
Selfie
Match scoreYour threshold
- Face found on the document yes
- One face on the selfie yes
- Document portrait reused no
Match Above the threshold
For developers
The score in the result.
The face_match check is part of every verification, with its status, its score and the evidence used.
Request
curl https://verify.atlasidv.com/v1/verifications/7f3c9a2e-… \
-H "Authorization: Bearer atlas_sk_sandbox_…" Response 200 OK
{
"checks": [
{
"type": "face_match",
"status": "passed",
"evidence_refs": [
"evidence:id_front",
"evidence:selfie"
]
}
]
} Take the next step
Try Atlas Face on your own flows.
A demo on your real cases, then a sandbox space with synthetic identities to integrate without any real data.