Enrolling Photos
Enrolment is what makes a person recognisable. A subject with one poor photograph will be missed; one with several good photographs will be found reliably.
Adding a photograph
Add photos from the subject's profile, or enrol somebody new from a face search that found nothing.
Where a photograph contains several faces, eConnect selects the largest and highest-quality one rather than rejecting the image — so a group photo can still be used, provided the person you want is the prominent face in it.
What makes a good enrolment photo
| Better | Worse |
|---|---|
| Face square to the camera | Steep angle or profile |
| Even lighting | Heavy shadow, backlight |
| In focus | Motion blur |
| Neutral expression | Extreme expression |
| Nothing obscuring the face | Sunglasses, mask, hand |
Several photographs beat one good one
The most useful thing you can do for recognition accuracy.
People do not present to cameras the way they present to an enrolment photo. Several images — different angles, different lighting, with and without glasses, across seasons — cover far more of the ways a person will actually appear.
Two or three good photographs materially outperform one excellent one.
Close matches when enrolling
If a photograph you are enrolling closely matches an existing subject, eConnect warns rather than silently creating a second record. That is the mechanism preventing duplicates, and it is worth reading rather than clicking past.
Where the match is genuinely the same person, add the photo to the existing subject rather than creating a new one. Where it is a genuine lookalike, proceed — but check first, because merging two records later is more work than getting it right now.
When a subject has many photographs
A profile built up over years can carry a great many enrolment photographs. They are all shown, and loading them is quick — if a profile takes a noticeable moment, that is the images arriving rather than anything being wrong.
Removing a photograph
Photographs can be unenrolled where one is harming rather than helping — a mislabelled image, or one so poor it produces false matches.
Removing a photo does not remove the subject or their detection history.
Splitting and merging
Where photographs of two different people have ended up on one subject, they can be split apart; where one person has been enrolled twice, the records can be merged. See Similar Faces and Merging.
Bulk enrolment
For enrolling many people at once — from a folder or an archive of images with accompanying data — use the bulk enrolment tools in Module Settings.
Audit record
What eConnect writes to the audit trail for the actions on this page.
SEARCH PHOTO; {context}{length} FACES RETURNED{known part}