[gsoc19-faces-workflow] Face management software comparison

[woenx]

I think it would be interesting to compare how other photography software
deal with face detection and recognition.

I found a few examples:

Google Photos (cloud)
https://www.youtube.com/watch?v=BbxmP0hT1iw

This is maybe the most deviant solution of the group, since it's a cloud
service rather than a desktop software. However, it is one of the most
simple ones, and I think it is a good example. Face recognition is performed
automatically on your uploaded photos. There's an "Album" view where the
interface shows a mix of events, folders and other dynamic elements. Among
these, you can find People, Places, Things, etc. where I assume google used
the metadata together with AI to create these classifications. The "People"
section is the one that is interesting to us. Faces are automatically
grouped, and you are only asked for their name. Future faces of that person
will automatically be tagged. If some specific face cannot be recognized,
you can still enter the name manually. At least in the Android version, you
can pick the name from your address book. This prevents spelling mistakes in
the names (which can easily lead to duplicated people). Sometimes faces are
detected incorrectly, but I have not found a way of correcting individual
faces. As a fun note, it also recognises pictures of pets, not only people.


Picasa
https://www.youtube.com/watch?v=9EvRboZeQIk

Google's previous project for a picture library, and now deprecated. The
process is not as straightforward, but it is very intuitive and offers more
options. If face recognition is turned on in the settings (you can select
which folders will be scanned for faces and which won't), Picasa will start
detecting faces right away as pictures are scanned. A category for "Unknown"
faces will be quickly populated, and similar faces will be grouped together
automatically. Selecting one of these groups shows all the containing faces
in a side panel. Selecting any single picture (or folder) also uses that
side panel to show which people are in it. Groups can also be expanded, so
all faces are shown. 
Once a person has been tagged (selecting a name from your Google contacts),
other faces belonging to that person are recognised and tagged
automatically. If Picasa is not sure, it will ask for confirmation, adding
that face to the end of the person's face list, and showing an Orange
asterisk right to their name. 
Sometimes Picasa seems to pause face recognition until the user intervenes,
and won't continue until more faces are manually recognized, which can be a
bit annoying at times, as it is often a lenghthy process that should require
minimal user intervention. 
Another feature is the ability to ignore faces, which are then hidden within
the "Unknown" section, and visible by clicking a button in that section.

BTW, Picasa is able to save XMP metadata to the pictures, but is not able to
re-read those faces unless the library is manually rescanned.


Lightroom CC
https://helpx.adobe.com/lightroom-classic/help/face-recognition.html (video)

In Lightroom, the process is similar, but more limited. By going to the
"People" section, faces are automatically scanned, detected, and placed in a
"Unnamed People" section. This can be done for the whole library, or on a
folder to folder basis. Similar faces are grouped on the go, showing the
number of faces for each group. Clicking on that number you can expand and
contract each group. Therefore, the "People" section in Lightroom works more
or less like Picasa's sidepanel, as a place to see what people are present
in a specific folder/album/picture.

However, Lightroom does not seem to automatically recognize and tag faces,
going only as far as suggesting the name of each faces in the "Unnamed
People". Therefore, it mainly offers face detection. Tagging a selection of
faces at once is possible. Face rectangles cannot be manually ignored, just
deleted.

People tags (keywords) are not mixed with the rest of keywords, and are
listed separately in a side panel, similar to the "Captions" sidepanel on
Digikam.


Mylio:
https://mylio.com/support/face-recognition/

I have personally not tried this software, but the face-tagging process is
very well documented on their website. 

Detecting faces in photos can be done picture by picture, or for the whole
library at once. Similar to other programs, Mylio will group similar faces
and ask for the user to name that person. Mylio will suggest a name ask for
confirmation when it thinks a new face belongs to a person. It also has an
option to ignore faces, but I am unsure what is the exact mechanism (if they
are deleted or just hidden).

Contrary to other software, if I am not mistaken, there is not a global
"Unknown" category for faces, but instead each album will show the number of
unidentified faces that contains.


Fotobounce
http://fotobounce.com/
https://www.youtube.com/watch?v=ureNGKrQC3I

I have never tested this software either. Similar to other programs, it
features face detection, and face recognition suggesting the name of the
person, but apparently does not automatically tags recognised people (like
Lightroom and Mylio). It also features an "Unidentified faces" section that
groups similar faces (where you can accept the suggestion all at once, like
in Picasa and Lightroom). 


Microsoft Photos
https://www.windowscentral.com/windows-10-photos-app-finally-adds-option-tag-people

This is the photo manager that comes with Windows 10. I haven't used
extensively, but it automatically scans your picture library in the user's
Images folder (if you activate that option in the settings). It includes a
"People" section where faces can be detected and tagged. In that regard, it
is similar to Google Photos, in which other objects and places, not just
faces, are recognized.




--
Sent from: http://digikam.1695700.n4.nabble.com/digikam-devel-f1695701.html