Faces in photos

Shotwell supports tagging images using faces of persons and assigning a name to them. If enabled, Shotwell also supports automatic detection of faces in images and recognition of already known faces in new images

Manual creation and modification of named faces

  1. Double-click on an image in the collection, then press the Faces button on the toolbar

  2. Start dragging a rectangle around the face you are tagging using your left mouse button

    You can also drag the rectangle by clicking into it or resize it by clicking on one of the borders

  3. Once satisfied, enter a name in the text box below the rectangle and press Enter to assign the name to the marked face

  4. Repeat this process for as many faces as you want to mark

  5. After you marked all faces, click on Ok to save the marked faces.

Modifying or removing existing named faces

An existing named face can be either be removed from an image using Faces ▸ Remove Face from Photo, by pressing Ctrl+R while in the face browser or by clicking Delete in the Faces tool

To rename the face, use Faces ▸ Rename Face... or press Ctrl+E or by clicking on Edit in the Faces tool.

Automatic detection of faces

It is possible to use automatic face detection to simplify the process of marking the faces. To trigger the face detection algorithm, click on Detect faces...

If reference images for faces are defined, this will also try to match any found faces in the current image with known references.

Automatic recognition of faces

A named face on a photo can be set as a reference for this particular name. Future calls to the automatic face detection will try to associate similar faces the same name if considered matching.

To declare a face as the reference for its name use these steps

  1. Switch to the face browser by clicking on a face name in the navigation tree.

  2. Select an image that contains the representation of the face you want to make a reference

  3. Click on Faces ▸ Train Face from Photo

Improving face recognition

By default, Shotwell uses cascade classifier to find faces in images. There is a method to use a deep neural network to improve detection rates significantly, but due to certain uncertainties with the licensing of the trained model it is not included by default.

To enable DNN-based face detection, download the res10_300x300_ssd_iter_140000_fp16.caffemodel from the OpenCV third-party repository

Copy the file into your local data folder, usually $HOME/.local/share/shotwell and restart Shotwell