Build a family-photo selection with digiKam’s face suggestions

A birthday slideshow needs photographs of the right person. digiKam’s People view suggests matches after you label reference faces; reviewing those suggestions gives you a selection to choose from.[4]

You will need desktop digiKam 9.1.0, a photograph collection already added to it and an album containing people you can identify. Begin with photographs that have not yet been scanned for faces. The stable release is available for Windows, Linux and macOS. This walkthrough follows project documentation; we have not run a photograph collection through it.[1][2]

Give the first pass a manageable boundary

Start with one event or one small album. Choose a person who appears in several clear photographs and whose identity you know. That gives the first review an answer you can check. A folder full of distant faces and unfamiliar guests is a harder place to learn what the suggestions mean.

Face analysis requires downloaded model files. If you skipped them at first launch, open Settings → Configure digiKam → Miscellaneous → System and use Download Required Binary Data. In Miscellaneous → Behavior, leave Enable background face recognition scan off for this exercise. That option starts another recognition pass when you tag a new face; manual passes make the initial sequence easier to follow.[3]

1. Open People in the left sidebar. In Search-in, limit the scan to your chosen album. Choose Scan new images in Settings, then Scan collection for faces.
2. Open Unknown. For a clear face, enter a name in Who Is This? and choose Confirm. Repeat with a few reliable examples.
3. Keep the album selected in Search-in. Choose Recognize faces only, then scan again.
4. Open Unconfirmed to inspect proposed names. Review each face before confirming it.

Use the face-name field: the general Assign Tag command labels the photograph. People thumbnails represent individual faces, so group photographs can appear repeatedly.[4]

Turn suggestions into decisions

Another suggested name can appear after a rejection. The project changed this behaviour in 8.7: it records rejected matches and offers the next candidate when one exists. Rejecting a wrong name is therefore different from identifying the right person. If you know who it is, supply that name and confirm the correction.[5]

There is no need to empty every unresolved queue before making a useful selection. An uncertain face can wait. For this first album, concentrate on the person you chose and inspect the photograph itself whenever the cropped face is ambiguous.

For another pass, use recognition again. The Maintenance tool’s Reset and clear all faces and training option restarts the face database and returns confirmed faces to Unknown. Its separate Rebuild all training data option reconstructs training from existing face thumbnails. Neither is an ordinary next step in this small review exercise.[7]

Make the selection useful outside digiKam

Open the person’s People entry and choose the photographs you want, checking any remaining suggested names before including them. Then use Export → Export to local storage. The export dialog starts with the selected items; inspect that list so you know which files will leave the collection.[6]

Create a fresh destination folder, choose Copy files and leave Overwrite existing items off. Copying produces files you can move elsewhere; symbolic links still point back to the originals. Leave Adjust Image Properties off if this selection should retain its existing format and dimensions. If you want copies without embedded Exif, IPTC or XMP data, use Remove all Metadata and leave sidecar copying off.[6]

Open the destination folder and inspect the saved photographs before sharing or printing them. Check that the intended person appears in each and that the selection contains the moments you want. You now have a clear stopping point for the first album: chosen pictures, checked identities and files ready for the next use.

Report an error or suggest a correction