Allow bulk confirmation of multiple different face identities across selected images

Brainstorm
1

Description

Currently, the face recognition workflow in digiKam requires confirming faces per person (tag), even when multiple faces across multiple images are already correctly identified.

While it is possible to:

  • select multiple faces and confirm them for one person

  • batch confirm similar faces grouped by similarity

…it is not possible to confirm multiple different identities in one action.

Problem

When working with images that contain multiple people (e.g. family photos, group shots), the workflow becomes inefficient:

Example:

  • 6 images

  • 3 people per image

  • all faces correctly recognized

Current workflow:

  • confirm Person A (all images)

  • confirm Person B (all images)

  • confirm Person C (all images)

:backhand_index_pointing_right: requires 3 separate actions instead of 1

This becomes very time-consuming for large collections (thousands of faces), as also noted by users struggling with bulk confirmation workflows

1 Like
2

In the Unconfirmed items you can do it. Somewhere else you suggest that could be possible?

Screenshot_20260415_145944
Screenshot_20260415_1459441872×1032 311 KB

3

According to your picture: I’ve tried this and all selected photos are assigned same name: Yane, or am I missing something. I am using digikam beta version 9.1.0 from 20260415. BR, Marjan

4

Yes, it is what is happening for now selecting faces of different people. Your suggestion is valid, I believe. Even that it costs one or two clicks more.

5

I would even go one step further:

Why is confirming a person necessary at all?

Or, more precisely: why should it be required before the recognition results become useful for searching?

I have a large photo collection (tens of thousands of images). The new face recognition works surprisingly well. After training with only a few confirmed faces, digiKam correctly identifies many more faces automatically.

For example:

  • Person X has more than 1000 detected faces.

  • In the People view, digiKam shows both confirmed and unconfirmed faces for X.

  • Most of the unconfirmed faces are actually correct.

However, when I try to search for photos containing person X (or combinations such as X AND Y), it seems that only confirmed faces are taken into account.

This makes me wonder about the intended workflow.

If digiKam has already recognized a face with high confidence and assigned it to person X, why can’t I use these results for searching?

From my point of view, this would be much more useful:

  • Use confirmed + unconfirmed matches for searching (perhaps as an optional mode).

  • Continue using confirmed faces only for training the recognition model.

For me, a false positive in a search is not a serious problem. If one photo of person Z appears in the search results for person X, I immediately see the mistake and can correct it.

A false negative is much worse: a correctly recognized photo never appears in the search results at all.

In other words, I would prefer a search mode that says:

“Show me everything that digiKam believes contains X (and optionally Y), even if some of those matches are still unconfirmed.”

This would make the face recognition much more useful for large collections, without requiring users to manually confirm thousands of already correct detections.

Is there a technical reason why unconfirmed recognitions are excluded from the person search, or is this simply the currently intended workflow?