<div dir="ltr"><div>Thank you for sharing these very thoughtful and clearly-phrased questions.</div><div><br></div><div>I don't know any of the answers, but if anyone can share insights on even a few of them it would be helpful to me as well.</div><div><br></div><div>There is some helpful information about face tags with rotated images (that may apply to cropped images as well) in this tutorial:<br></div><div> <a href="https://userbase.kde.org/Digikam/Tutorials/Tagging_and_Face_Tags#Rotating_face_tags">https://userbase.kde.org/Digikam/Tutorials/Tagging_and_Face_Tags#Rotating_face_tags</a></div><div><br></div><div>About me:</div><div>I've recently transitioned our 20+ years collection of about 190,000 images from Picasa to DigiKam, and am in the process of cleaning up metadata as well. I am currently using DigiKam on multiple Windows laptops to manage a shared database and image collection on a NAS.<br></div><div><br></div><div>Best regards,<br></div><div>Will.<br></div><div><br></div><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sun, Oct 16, 2022 at 5:57 AM Jens Benecke <<a href="mailto:jens-digikam@spamfreemail.de">jens-digikam@spamfreemail.de</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hello,<br>
<br>
I am currently cleaning out my photos database of ~150'000 images, taken <br>
in the last 20 years roughly.<br>
<br>
Until 2015, this database was maintained using iPhoto, and I converted <br>
this to Digikam using a Ruby script (on Github) and a lot of time <br>
manually correcting stuff. I noticed that some face rectangles were at <br>
an incorrect position, especially with cropped or rotated images, and a <br>
lot of labels were simply wrong.<br>
<br>
To optimize my future workflow and improve face detection, I would like <br>
to know a bit more about how Digikams face detection actually works. So <br>
I have these questions:<br>
<br>
1. How big does a face rectangle need to be if I draw one manually? Does <br>
it need to contain the whole head of the person, or is "eyes + nose + <br>
mouth" sufficient? Does this size need to be consistent throughout one <br>
person or does Digikam not care at all?<br>
<br>
2. How tolerant is Digikam regarding rotated faces, e.g. when people are <br>
lying, upside down, looking sideways (= profile view), or looking down <br>
(skewed view)? Does Digikam regard these faces as "different"? If so, is <br>
this avoidable? I noticed that after importing a lot of images with one <br>
person lying down, Digikam would suddenly assign this person to ALL <br>
photos where ANYbody was lying down.<br>
<br>
3. How tolerant is Digikam regarding "decoration" like glasses, scarves, <br>
half-obscured faces etc? Do such images pollute the learning process?<br>
<br>
4. Does Digikam care about exposure, brightness, or sharpness of images? <br>
Do these images "hurt" the faces learning process?<br>
In other words, can I manually tag images where faces are not detected <br>
and maybe not even clearly visible, but where I *know* the person is <br>
there, just for my organization, or will this poison the learning database?<br>
<br>
5. How tolearnt is Digikam regarding false positives? If 1 of 100 images <br>
for a person is incorrect (but confirmed), how much will this interfere <br>
with the detection of future faces?<br>
<br>
I am currently using Digikam 7.9.0 appimage 2022-10-12, on KDE Neon.<br>
<br>
<br>
Thank you!<br>
<br>
-- <br>
Regards, Jens<br>
<br>
</blockquote></div></div>