How to Identify Closed Eyes in Wedding Photos

Why blinks are the sneakiest cull failure

A blink is invisible at thumbnail size and glaring in a delivered print. The worst case is the family formal: with twelve people in frame, the odds that someone is mid-blink in any given shot are high, which is why experienced photographers shoot each grouping several times. But that habit moves the problem to the cull, where you must now compare five nearly identical group shots face by face. Miss one and it is exactly the photo the couple's mother orders framed; the problem page on closed eyes ruining group photos goes deeper on the stakes.

A concrete example

Say you shot six frames of a ten-person family grouping. Checking every face in every frame is sixty face inspections for one delivered photo. The efficient order is: pick the frame with the best overall expressions at a glance, then sweep its faces left to right at 100%. If one uncle is blinking, check only that uncle in the other five frames to find a swap candidate or a replacement keeper. That is closer to fifteen inspections than sixty.

Step-by-step blink check

1. Batch the formals together so all takes of each grouping sit side by side.
2. Pick a lead candidate per grouping on overall expression and composition.
3. Sweep the candidate's faces at 100%, one direction, no skipping, including the people at the edges.
4. On a blink, compare just that person across the other takes and either switch keepers or note a head-swap for editing.
5. For portrait bursts, check eyes only on your winner; for candids, reject only when the blink belongs to the subject of the frame.

Limitations

Automated eye detection needs a detectable face at a reasonable size: profiles, faces tiny in a wide shot, heavy backlight, veils, and sunglasses all reduce reliability, and squinting laughter can read as closed eyes when it is actually the best frame emotionally. Human review of group formals is still worth the minutes even with automated flagging, precisely because the cost of one miss is so high.

How Cull AI Studio handles closed eyes

Cull AI Studio runs face detection and open/closed-eye analysis on every frame, locally and offline, using a bundled face detector and eye classifiers, and factors the result into each KEEP, REJECT, or REVIEW decision. Frames where eyes appear closed on detected faces are flagged without you zooming into anything, and low-confidence calls, such as small or partly turned faces, are routed to REVIEW rather than quietly decided. Closed-eye corrections you make are recorded by reason, and in the staged "Sarah & Dimitri" demo report, closed eyes was the most common correction reason at 12 of 31, which is demo data but matches how weddings go. See also can AI detect closed eyes in photos.