Announcement

Collapse
No announcement yet.

Face Recognition Questions

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Face Recognition Questions

    I hope these haven't already been covered, searched but didn't find exact answers. Face recognition is a really appreciated addition and upgrade from my 2018 Pro version and I spent several ours with my new program last night entering them. I did find a lot of faces not automatically entered after running cataloging, even though I had created several images of each person.

    I have the following questions:

    Where is the face information stored, just in the ACDSee data base, or in side car files, or? I could't find any names in the IPTC or EXIF or the Metadata list. It would be nice if the names would show in the Metadata list also.

    What files are the face data image compatible with if a file is converted? Do they only show up when viewing with ACDSee 2019? I assume so, because I opened up the same files in 2018 and there is no indication when I re catalog the same files that were recognized in 2019.

    Is there any way to select numerous files with face names and change the names? That would be a nice addition.

    Once you have a name in the Catalog - People list and no longer are using that name, how do you remove it from the list??

    Doesn't seem to work as well on animals, maybe I'm asking for too much. :-)

    Thanks,




  • #2
    Just discovered those controls for people under the Tools. I can now easily rename people or delete them as well.

    Comment


    • #3
      Where is the face recognition info saved.. just in the ACDSee Database? I noticed when I made a bunch of face changes it didn't show the Embed icon on the files and at close it didn't embed those changed files. Is face recognition not embedded with the files or the side car files?

      Anyone know? It looks like it's just in the ACDSee Database, which means it is probably really important to keep backups frequently and copies of backups?






      Last edited by doane; 03-17-2019, 05:31 AM.

      Comment


      • #4
        No answers, so I did some experimenting. If a file is given face info and metadata (caption, rating, etc.) and it is copied to another folder using ACDSee metadata and face tags are retained as it is if the metadata is embeded first.
        If you convert a file with metadata and face tags to a different format, the metadata is retained, but the face info is lost. ie: I converted a psd file to jpg and checked the boxes for both the preserve metadata and the preserve database info and the face tags are lost.
        So, for some reason, the face tag info is not being preserved when files are converted to a different format.. this is not good.
        I have a lot of psd files and tiff files that I frequently convert to jpg for internet or emails, etc. Obviously the face tags can't be seen unless viewing in ACDSee, but when managing files with the program I don't want to have to redo the face tags

        Comment


        • #5
          Hi doane,

          All face detection and recognition data is stored in the database and can't be embedded in images. This is a feature we have seen a lot of feedback about, and it is something we would like to add in the future. We will also keep the file format conversion case in mind. The metadata and face data you are noticing transferring from location to location occurs because ACDSee copies database information automatically when files are copied and pasted.

          Thanks,
          Tristan H.
          ACD Systems

          Comment


          • #6
            Thanks, Tristan: I can understand the face data not being embedded, would be nice if they could, but for the face info to be lost when converted by ACDSee to a different file format? That's more serious and something that it seems like should be addressed soon.

            Comment


            • #7
              Hi doane,

              Converting a file to another format means changing the color values of every single pixel, and often involves changing the bit depth as well. To a human the new file looks almost identical, but if face detection is run on the new image, the results will be different. Generally speaking, the results will be similar enough that the end result (detection and recognition of the person) is the same. But to simply copy the old database records over to the new file would mean copying inaccurate information.

              Thank you for mentioning this case. We will have to decide what the most optimal solution is, whether it would be better to accept slightly inaccurate information to save time, or redo detection to ensure maximum accuracy. As you say, saving time is a real concern for people who convert lots of files.

              Thanks,
              Tristan H.
              ACD Systems

              Comment


              • #8
                > Converting a file to another format means changing the color values of every single pixel...
                > if face detection is run on the new image, the results will be different.
                > ... simply copy the old database records over to the new file would mean copying inaccurate information.

                Madness. It's a copy. It has the same logical content as the original, even though the pixels maybe different. If it was a picture of Bob before the conversion, it's still a picture of Bob after the conversion. There is no need for ACDSee to rerun the face detection on the copied image, and introduce the risk of failing to detect Bob.

                "redo detection" will not "ensure maximum accuracy"; it only gives ACDSee the opportunity to get it wrong.

                Comment

                Working...
                X