Updated October 18, 2026: Replaced the old sensitivity-only workflow with the current detection, training and review process.
digiKam’s People view can help organize a photo collection by person. The useful workflow is to detect faces, label a representative set correctly, then review recognition suggestions. Automatically finding a face is not the same as reliably knowing whose face it is.
Prepare a small test collection
Install digiKam from its official project source or your operating system’s supported package source. Back up your photo collection and digiKam database before a large reorganizing job. Start with one album containing clear examples instead of scanning every disk immediately. If the pictures came from Google Photos, verify the account transfer and saved copies before reorganizing them.
The current digiKam manual says face management needs downloadable deep-learning model files. If the application requests those components, use its supported download option. Missing models should be checked before repeatedly rescanning an unchanged collection.
Detect faces first
Open People in the left sidebar. Use the search-scope controls to choose the albums or tags you want to scan, and inspect the Settings tab. Start with the option for new/unscanned images when that matches your goal; rereading an entire collection can take much longer.
Detection finds face regions. Review missed faces or incorrect detections, especially in group photographs, profiles and difficult lighting. Add or correct a face region manually when necessary.
Label examples carefully
Assign the intended person’s name to several clear, varied examples. Avoid training on an uncertain suggestion just because the software ranked it first. A mislabeled person can create more correction work later.
Use consistent names so that the same person does not end up under several near-duplicate tags. For children, appearance changes or difficult angles, expect to confirm additional examples instead of relying on a single photograph.
Run recognition and review its suggestions
Recognition uses the confirmed examples to suggest identities for other detected faces. Review the suggested matches and accept, reject or correct them. Similar-looking relatives and older images deserve particular attention.
Do not assume a higher numeric slider always means the same thing across versions or controls. Read the setting’s current description: detection sensitivity and recognition confidence address different decisions. Test a setting on the small album before applying it to the whole library.
Protect the catalog and the people in it
Back up the database and check which metadata digiKam writes into image files or sidecars. Before sharing exported pictures, inspect whether names or other personal tags are included. Organizing locally does not mean every future export will automatically omit that metadata.
For a large library, plan time and disk space for scanning and review. Keep originals until you have checked the resulting catalog and backup. Face recognition is an organizing aid, not a reason to delete unrecognized photos.
Images from the original article
These original illustrations are retained for reference. Product examples and software screens may show older versions; use the model-specific instructions above.
