Engine · Vision library

People detection

Detect faces, build person groups and keep corrections attached to the source photos.

Workflow

1. Select photos for people analysisChoose eligible sources under the product consent and policy boundary.
2. Detect and group facesProduce source-linked face evidence and groups; missing faces and abstention remain valid outcomes.
3. Review uncertain resultUncertain, incomplete or user-sensitive results remain reviewable with their source identity and reasons.
4. Show people groupsPresent groups with photo counts and correction actions without claiming an unknown identity.
5. Return unavailableMissing prerequisites and unsupported inputs produce an explicit unavailable outcome without inventing a result.

Failure boundaries

• No consentNo analysis: Do not start model
• Face without sourceInvalid evidence: Hide group
• Wrong groupingReview item: Correct membership

engine.vision.people