<html>
  <head>
    <meta http-equiv="content-type" content="text/html; charset=UTF-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <div dir="ltr">Hello there,<br>
      it has been a while since the last activity in this thread but
      I've installed digiKam 6 beta this week and checked if I still
      have the same troubles.<br>
      Before I start I may clarify a few things and would like to make
      some suggestion to ease the face tag / other tags discussion.</div>
    <div dir="ltr"><br>
    </div>
    <div dir="ltr">Prior this, I have to mention I try to avoid to use
      the word <i>tag</i>, as it used inflationary. All metadata
      records are stored in fields (see e.g. <a moz-do-not-send="true"
href="http://photometadata.org/META-Resources-Field-Guide-to-Metadata#Keywords">photometadata.org</a>)
      which also often called tags (of the metadata), so a tag is
      anything what is used in digiKam to filter or search for images,
      e.g. keywords, colour label, star rating etc... Thus there  is to
      much space for interpretation what leads to all these questions
      due to irritations caused by the use of the word tag.
      <div dir="ltr">In order to lower the entry hurdle into the world
        of tagging I would suggest to be consistent with the <i>offical
        </i>wording, thus new users won't be confused by this. That
        means that the text for the tag will be named keyword, so on the
        source selection pane on the left will be <i>Keywords</i> and
        in filter pane on the right it will say Keywords Filter. The
        description shall rather say close to <i>digiKam deals with
          metadata, grouped in (tags of): keywords, label, date and
          location</i>. </div>
      <div dir="ltr"> <br>
        First of all, Scan for Faces lists the results in the people
        view windows, see digiKam docs, where the user can confirm or
        correct the assumed names of the algorithm of the person in the
        image. <br>
        So far so good. As soon as a name is confirmed digiKam writes
        the data to the MP and MWG namespace of the XMP records, it sets
        a name and area.</div>
      <div dir="ltr">
        <div dir="ltr">More Details about those namespaces can be found
          here:</div>
        <div dir="ltr">
          <ul>
            <li><font size="-1"><a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/Microsoft.html#MP"
                  moz-do-not-send="true">Microsoft MP Tags (ExifTool)</a>
                <br>
              </font> </li>
            <font size="-1"> </font>
            <li><font size="-1"><a moz-do-not-send="true"
                  href="http://www.exiv2.org/tags-xmp-MP.html">Microsoft
                  MP Tags (Exif2.org)</a></font></li>
            <font size="-1"> </font>
            <li><font size="-1"><a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/MWG.html#Regions"
                  moz-do-not-send="true">MWG Regions Tags (ExifTool)</a>
                <br>
              </font> </li>
            <font size="-1"> </font>
            <li><font size="-1"><a moz-do-not-send="true"
                  href="http://www.exiv2.org/tags-xmp-mwg-rs.html">MWG
                  Regions Tags (Exif2.org)</a></font> </li>
          </ul>
          <p>as Apple and Adobe write their information in the MWG
            namespace, it I would say that MWG is the leading namespace
            but inconsistent may lead to unexpected behaviour of the
            applications what reads them. <br>
          </p>
          <p>In my understanding these information should also be
            written to the <a
href="https://iptc.org/std/photometadata/specification/IPTC-PhotoMetadata#person-structure"
              moz-do-not-send="true">IPTC Person structure</a> as
            mentioned in the <a
href="https://www.iptc.org/std/photometadata/documentation/userguide/index.htm#!Documents/personsdepictedintheimage.htm"
              moz-do-not-send="true">IPTC Photo Metadata User Guide
              (Persons Depicted in the Image)</a>, but is not. May
            someone can explain why?<br>
          </p>
        </div>
      </div>
      <div dir="ltr">In order to make images findable by a person's
        name, the name shall also be written to the keywords field of
        multiple namespaces, <a
href="https://www.iptc.org/std/photometadata/documentation/userguide/index.htm#!Documents/personsdepictedintheimage.htm"
          moz-do-not-send="true">IPTC Photo Metadata User Guide (Persons
          Depicted in the Image)</a> recommends caption and keywords. I
        cannot tell all relevant fields/namespaces. My research tells me
        that
        should be at least those: <br>
      </div>
      <div dir="ltr">
        <ul>
          <ul>
          </ul>
        </ul>
        <ul>
          <li>IPTC ⇒ Keywords, see: <a
href="https://iptc.org/std/photometadata/specification/IPTC-PhotoMetadata#keywords"
              moz-do-not-send="true">iptc.org</a></li>
          <li>XMP acdsee (<a moz-do-not-send="true"
              href="https://www.acdsystems.com/">ACD Systems</a>) Tags ⇒
            catergories, see: <a
              href="http://www.exiv2.org/tags-xmp-acdsee.html">ExifTool</a>
            / <a href="http://www.exiv2.org/tags-xmp-acdsee.html">Exif2.org</a></li>
          <li>XMP digiKam Tags ⇒ Tag List, see: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#digiKam"
              moz-do-not-send="true">ExifTool</a> / <a
              href="http://www.exiv2.org/tags-xmp-digiKam.html">Exif2.org</a></li>
          <li>XMP dc (<a moz-do-not-send="true"
              href="http://dublincore.org/">Dublin Core</a>) Tags ⇒
            Subject, see: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#dc"
              moz-do-not-send="true">ExifTool</a> / <a
              href="http://www.exiv2.org/tags-xmp-dc.html">Exif2.org</a></li>
          <li>XMP lr (Lightroom) Tags ⇒ hierarchicalSubject, see: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#Lightroom"
              moz-do-not-send="true">ExifTool</a> / <a
              moz-do-not-send="true"
              href="http://www.exiv2.org/tags-xmp-lr.html">Exif2.org</a><br>
          </li>
          <li>XMP <a moz-do-not-send="true" href="MediaPro">MediaPro </a>Tags
            ⇒ Catalog Sets, see: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#MediaPro"
              moz-do-not-send="true">ExifTool</a> / <a
              moz-do-not-send="true"
              href="http://www.exiv2.org/tags-xmp-mediapro.html">Exif2.org</a><br>
          </li>
          <li><a moz-do-not-send="true"
href="https://docs.microsoft.com/en-us/windows/desktop/wic/-wic-codec-metadatahandlers">Microsoft
              XMP</a> Tags ⇒Las Keyword XMP, see: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/Microsoft.html#XMP"
              moz-do-not-send="true">ExifTool</a> / <a
              moz-do-not-send="true"
              href="http://www.exiv2.org/tags-xmp-MP.html">Exif2.org</a></li>
          <li><a moz-do-not-send="true" href="metadataworkinggroup.com">MWG</a>
            Keywords Tags: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/MWG.html#Keywords"
              moz-do-not-send="true">ExifTool</a> / <a
              moz-do-not-send="true"
              href="http://www.exiv2.org/tags-xmp-mwg-kw.html">Exif2.org</a></li>
        </ul>
         has a field but ignored by digiKam, why?</div>
      <div dir="ltr">
        <ul>
          <li>XMP <a moz-do-not-send="true"
              href="http://www.prismstandard.org/">prism</a> Tags ⇒
            Keyword: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#prism"
              moz-do-not-send="true">ExifTool</a> <br>
          </li>
        </ul>
        but to be exclude are:<br>
        <ul>
          <li>XMP <a moz-do-not-send="true"
              href="https://www.adobe.com/devnet/xmp.html">xmp</a> Tags,
            as it says <i>non-standard</i>: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#xmp"
              moz-do-not-send="true">ExifTool</a> <br>
          </li>
          <li>XMP xmpMM Tags as it says <i>undocumented</i>: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#xmpMM"
              moz-do-not-send="true">ExifTool</a> <br>
          </li>
          <li>XMP pdf Tags: <a
href="https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/XMP.html#pdf"
              moz-do-not-send="true">ExifTool</a>-> only for Adobe
            PDF <br>
          </li>
        </ul>
      </div>
      <div dir="ltr">I reckon there isn't any leading field as a
        mismatch could lead to inconsistent search result, depending
        heavily on the used application.<br>
      </div>
      <div dir="ltr"> <br>
      </div>
      <div dir="ltr"> I mentioned that, since as far as I know that
        content of person related metadata fields are not taken into
        account when you search or filter a collection by certain
        keywords. Thus, in order to make the names findable by digiKam,
        the name has to be added to the keywords related metadata fields
        to make the magic happen.</div>
      <div dir="ltr"><br>
      </div>
      <div dir="ltr">If I'm correct, what is the source of the list of
        the people pane on the left? In my opinion there are three
        options. <br>
        <ol>
          <li> First, these are the keywords listed below the hierarchy
            level <i>persons</i> in the keyword list. If the user
            selects an name it filters images based on the keywords and
            shows the face area as described in the person related
            metadata field. </li>
          <li>Second, digiKam reads the information given in the person
            related fields of the metadata of each image in this
            particular case . Afterwards it uses this data to populate
            the person pane. That would be quite of workload on the CPU
            and isn't very likely. </li>
          <li>Third, it stores the information given in person related
            fields of the metadata of each image in the database
            recognition.db. Based on the information stored there
            digiKam knows which images are to be shown. In this case,
            are the face thumbnails are stored in this database as well
            or are they derived from each image, based on the region
            information?</li>
        </ol>
      </div>
      <div dir="ltr"><br>
      </div>
      <div dir="ltr">The last question in this regard. What happens to
        unknown/unconfirmed faces? Are they stored immediately to the
        database but metadata isn't written until face is confirmed?
        That would explain, why they appear again, when you open the
        left people pane, after restart of digiKam. In addition that
        would indicate that my third theory above is the right one.  <br>
      </div>
      <div dir="ltr"><br>
      </div>
      <div dir="ltr">In case the data in the left person pane isn't
        based on the keywords, as assumed before, the person tag shall
        be a classed differently, means it shall have a different
        symbol. That would help to distinguish easily between face tags
        and keyword tags. It would also the simplify the progress of
        understanding how tagging and face recognition work in digiKam.<br>
      </div>
    </div>
    <div dir="ltr"><br>
    </div>
    <div dir="ltr">In addition I would like to see some changes in
      regard to the <i>unkown </i>faces thumbnails.<br>
      <div dir="ltr">Those wishes are most likely discussed in other bug
        reports. For convenience I listed those created by woenx and
        mine again.</div>
      <div dir="ltr">As you see most wishes are still unresolved and
        mine will mostly a duplicate of presents ones I'll list them
        anyway in order to highlight their necessity. <br>
      </div>
      <div dir="ltr">I would like to be able to </div>
      <div dir="ltr">
        <ol>
          <li>stop auto refresh of the thumbnails to avoid confirming a
            wrong face accidentality. It is a pain in the arse to undo
            such accidents  </li>
          <li> sort them at least by guessed faced.<br>
            It would be preferred if sorting in any view is possible by
            any property what can be used to filter items</li>
          <li>drag and drop selected faces over an <i>Person Name</i></li>
          <li>assign <i>Person Name</i> via right click menu as
            possible for tags</li>
          <li>group similar faces in "Unknown" faces </li>
        </ol>
        <div dir="ltr">what do you think before I put new ones in wish
          reports?<br>
        </div>
        <div dir="ltr"><br>
          <ol>
            <a
href="https://bugs.kde.org/buglist.cgi?bug_status=__open__&component=Faces-Recognition&list_id=1583144&product=digikam"
              moz-do-not-send="true">digikam Bug List - Component:
              Faces-Recognition Status: REPORTED, CONFIRMED, ASSIGNED,
              REOPENED</a> <br>
            <a
href="https://bugs.kde.org/buglist.cgi?bug_status=__open__&component=Faces-Workflow&list_id=1583148&product=digikam"
              moz-do-not-send="true">digikam Bug List - Component:
              Faces-Workflow Status: REPORTED, CONFIRMED, ASSIGNED,
              REOPENED</a>
            <li>
              <p>Bugs</p>
              <ol>
                <li><del><a
                      href="https://bugs.kde.org/show_bug.cgi?id=392013"
                      moz-do-not-send="true">Bug 392013</a></del>:
                  Metadata explorer does not show XMP face rectangles</li>
                <li><del><a
                      href="https://bugs.kde.org/show_bug.cgi?id=392017"
                      moz-do-not-send="true">Bug 392017</a></del>:
                  Merging, renaming and removing face tags</li>
                <li><del><a
                      href="https://bugs.kde.org/show_bug.cgi?id=392009"
                      moz-do-not-send="true">Bug 392009</a></del>: Weird
                  automatic subtag within "Unknown people" called "da"</li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392008"
                    moz-do-not-send="true">Bug 392008</a>: Inconsistent
                  behaviour of "People" Tag</li>
              </ol>
            </li>
            <li>
              <p>Wishes</p>
              <ol>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=275671"
                    moz-do-not-send="true"><del>Wish 275671</del></a>:
                  Scan single image for faces </li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392015"
                    moz-do-not-send="true">Wish 392015</a>: Show
                  "Unknown" faces in a more visible and preeminent place
                  in the "People" list</li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392007"
                    moz-do-not-send="true">Wish 392007</a>: Face tags
                  and regular tags are mixed together and cannot be told
                  apart</li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392016"
                    moz-do-not-send="true">Wish 392016</a>: Confirmed
                  and unconfirmed faces look the same in a person's face
                  list</li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392020"
                    moz-do-not-send="true">Wish 392020</a>: No possible
                  way of knowing which pictures within a regular tag
                  have been face-tagged</li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392022"
                    moz-do-not-send="true">Wish 392022</a>: Position of
                  a face tag appears on top or bottom of the list,
                  instead of being sorted alphabetically</li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392023"
                    moz-do-not-send="true">Wish 392023</a>: Feature
                  request: add "Ignored" group of faces: </li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=392024"
                    moz-do-not-send="true">Wish 392024</a>: Feature
                  request: group similar faces in "Unknown" faces <br>
                  ⇒ <a
                    href="https://bugs.kde.org/show_bug.cgi?id=384396"
                    moz-do-not-send="true">Wish 384396</a>: Wish:
                  display faces sorted by similarity (pre-grouped)
                  instead of album/time/..</li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=386291"
                    moz-do-not-send="true">Wish 386291</a>: only refresh
                  found face list/pane upon user request </li>
                <li><a
                    href="https://bugs.kde.org/show_bug.cgi?id=254099"
                    moz-do-not-send="true">Wish 254099</a>: <span
                    id="summary_container"><span
                      id="short_desc_nonedit_display">SCAN : refresh
                      collection with a script in commandline</span></span></li>
              </ol>
            </li>
          </ol>
          <p> </p>
          <div bgcolor="#FFFFFF"><br>
            <br>
            <p><br>
            </p>
            <div class="gmail-m_9026140557679506384moz-cite-prefix">Am
              19.03.2018 um 06:16 schrieb Stefan Mueller:<br>
            </div>
            <blockquote type="cite">
              <div dir="auto">Great work 
                <div dir="auto">Will go through them as soon as I can
                  spare some time <br>
                  <br>
                  <div dir="auto">Sent from a fair mobile </div>
                </div>
              </div>
              <br>
              <div class="gmail_quote">
                <div dir="ltr">On Sun, 18 Mar 2018, 21:58 woenx, <<a
                    href="mailto:marcpalaus@hotmail.com" target="_blank"
                    moz-do-not-send="true">marcpalaus@hotmail.com</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">Ok, these are the
                  bug reports corresponding to face management bugs and<br>
                  improvements as discussed in this thread (and other
                  related threads), in<br>
                  case anyone wants to follow them:<br>
                  <br>
                  - Scan single image for faces: <a
                    href="https://bugs.kde.org/show_bug.cgi?id=275671"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=275671</a><br>
                  - Inconsistent behavior of "People" Tag:<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392008"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392008</a><br>
                  - Face recognition assigns faces to people without
                  confirmation, leading to<br>
                  false positives: <a
                    href="https://bugs.kde.org/show_bug.cgi?id=392010"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392010</a><br>
                  - Metadata explorer does not show XMP face rectangles:<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392013"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392013</a><br>
                  - Show "Unknown" faces in a more visible and
                  preeminent place in the<br>
                  "People" list: <a
                    href="https://bugs.kde.org/show_bug.cgi?id=392015"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392015</a><br>
                  - Face tags and regular tags are mixed together and
                  cannot be told apart:<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392007"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392007</a><br>
                  - Confirmed and unconfirmed faces look the same in a
                  person's face list:<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392016"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392016</a><br>
                  - Merging, renaming and removing face tags:<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392017"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392017</a><br>
                  - No possible way of knowing which pictures within a
                  regular tag have been<br>
                  face-tagged: <a
                    href="https://bugs.kde.org/show_bug.cgi?id=392020"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392020</a><br>
                  - Weird automatic subtag within "Unknown people"
                  called "da":<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392009"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392009</a><br>
                  - Position of a face tag appears on top or bottom of
                  the list, instead of<br>
                  being sorted alphabetically: <a
                    href="https://bugs.kde.org/show_bug.cgi?id=392022"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392022</a><br>
                  - Feature request: add "Ignored" group of faces:<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392023"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392023</a><br>
                  - Feature request: group similar faces in "Unknown"
                  faces:<br>
                  <a href="https://bugs.kde.org/show_bug.cgi?id=392024"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">https://bugs.kde.org/show_bug.cgi?id=392024</a><br>
                  <br>
                  PackElend, I didn't report the bug you explained in
                  your first post since I<br>
                  didn't fully understand the problem. I think it should
                  be better if you<br>
                  wrote a bug report yourself.<br>
                  <br>
                  And that's it. If all of these were solved, face
                  detection and recognition<br>
                  would be just perfect (apart from the matter of
                  accuracy, of course).<br>
                  <br>
                  <br>
                  <br>
                  --<br>
                  Sent from: <a
                    href="http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html"
                    rel="noreferrer noreferrer" target="_blank"
                    moz-do-not-send="true">http://digikam.1695700.n4.nabble.com/digikam-users-f1735189.html</a><br>
                </blockquote>
              </div>
            </blockquote>
          </div>
        </div>
      </div>
    </div>
  </body>
</html>