Acdsee 2020 How to embed face data in file?

[peter09]

I ran the face recognition. Under faces on the left in the panel I also see the names of the recognized faces.

Now I want the names in the file are given. For this I select the files, go to embed metadata and choose embed face data. This is also done correctly.
But then I see this face data (name) nowhere. Neither in IPTC, Exif, Acdsee metadata, etc.
I want the names contained in the exported file.
Is that possible.
What am I doing wrong?

[Tristan H]

Hi peter09,

We currently do not support viewing of ACDSee Face Data (in the database or embedded) in the Properties pane.

Thank you for the feedback though. I've passed it on to the development team, so it is something we can possibly explore in the future.

Thanks,
Tristan H.
ACD Systems

[Emil]

This has indeed be asked for several times in the past. Afaik all you can do is, select every single found face in the peoples section of the catalog pane, then select all items (CTRL-A) and add a proper keyword or assign the items to a category or collection with a suitable name.

If you want to export the images and give them to people who don't use AC, then imho IPTC-keywords are the best option.

[gmb]

I agree. I would like to have the ability to store the names detected in the IPTC description field for others to use.

[StaffsLebowski]

As a database professional (30 years, and counting..), I can tell you that retrieving this information is trivial, and a standard function of any database. You have an image, you related a name to that image - now the database can give you that information. This is week 1 stuff in the database world. If you can't access this information then it's because they don't want you to have it. Why? money of course. Because there's a chance you'd run off with your 150k+ images to another vendor. They have to lock you in somehow. Very poor..

[prsettle]

With respect, I doubt that the issue is to do with locking users in to the product or indeed whether or not it is possible to retrieve the data from the database, which I have no doubt is trivial. My understanding is (and I could well be wrong) that the issue was more to do with where to store the face data inside the image file in a way that is compliant with existing IPTC standards and therefore making it portable.

Up until December 2019 I don't think there was a standardized way of embedding face data into IPTC metadata. However, a quick google suggests that in December 2019 the IPTC standard was indeed updated to allow for such data to be embedded into the image files.

My guess is that ACD Systems are catching up with the latest standard that is only just over 1 year old. I'm sure we'll see it at some point...

[Greyfox]

.....Up until December 2019 I don't think there was a standardized way of embedding face data into IPTC metadata. However, a quick google suggests that in December 2019 the IPTC standard was indeed updated to allow for such data to be embedded into the image files.

As I understand it, the fields relating to PersonInImage are actually in Extension v1.5 to the standard, rather than in the IPTC Core v1.2, and are Xmp fields. In my experience uniformity across the various image applications when it comes to Xmp is not all that good. Whilst it would be nice to have the ability to embed the names into images in some stardardized format, its usefulness outside of ACDSee would be conditional on other software providing access to it, and as is currently the case, create issues if other software doesn't retain it.

Like many others, I use keywords for the names.

[stb62et]

Hello
nevertheless, i think the requirement from peter09 and emil are valid. Where is a place within this community where we can articulate our product improvements which we wish to see in the next release

[Regor250]

What data is being asked? Names of course, but face data would also contain location within the image of a face image frame and dimension of the frame for each face and a unique frame ID (frameID, x, y, width, height, name, status) as the same person's name can appear more than once in an image (mirror image, picture frames etc.). It is therefore not as simple as a single data field containing all the names associated with one image, it's a relational many to many data elements structure (many people with many data elements) whereas metadata information usually is one to many (where one field can contain many elements such as keywords). Therefore if all that is wanted is names added to the keywords field that's easy, if you want face recognition info (name, location etc.) then it's a wee bit more complex to manage as metadata goes. I can't think of any many to many data elements supported in image metadata structure and in all likelihood would not be widely supported, so what would be the benefit of that?

[StaffsLebowski]

Hi Prsettle,

<<the issue was more to do with where to store the face data inside the image file in a way that is compliant with existing IPTC standards and therefore making it portable.>>

You're presuming that a User wants to store face data in the IPTC. Storing this information in the IPTC has nothing to do with it. To what use, and where one decides to put the data, after it's been extracted - is up to the individual. The problem is, Users are not able to extract this information to do with as they wish. Therefore, I would definitely consider this as one of the ways a company tries to restrict Users from taking their photos and moving to another vendor.

Regards..

[Regor250]

What the OP is asking for is the ability to extract names from face data and store them where ever for some unstated purpose. Not certain how that differs from selecting files from a name and assigning that name as a keyword. The best ACDSee could do right now is to provide a function to automate that process.

To my knowledge there exists no Face Tagging Metadata standard (ISO XXXX), hence no single way to integrate that information into a different system without knowledge of proprietary data structure stored in ACDSee database and the same goes for any other systems on the market. ACDSee could create a metadata system within its own ACDSee Metadata, but it would still be proprietary and hence not particularly useful to others. Until international metadata standard bodies come up with a standard, there isn't much anything of use to others ACDSee can do.

[Greyfox]

Hello
nevertheless, i think the requirement from peter09 and emil are valid.

This thread was started quite some time ago and some of the content of the early posts is perhaps out of date.

The current version of ACDSee Photo Studio Ultimate 2021 DOES allow for face detection metadata to be embedded in the image, and the Tools->Database-Catalog Files option can restore the embedded face metadata into the images should that prove necessary.

The following screen shot shows the XMP face metadata embedded in an image by ACDSee U2021, as viewed in ExifToolsGUI



As can be seen, it not only includes the Name, but also the region information for the frame.

Both the face frame and the Name that goes with it can be seen in view mode. The name can also be seen in the Catalog Pane under People, and can be searched on either using Quick Search, or be selecting the People->Name in the Catalog Pane.

[Greyfox]

To add to my previous post, and in case anyone wonders what happens with multiple faces in an image, here is the face metadata saved in an image with three faces identified.

[StaffsLebowski]

<<What the OP is asking for is the ability to extract names from face data and store them where ever for some unstated purpose. Not certain how that differs from selecting files from a name and assigning that name as a keyword.>>

The difference is:

1: In the first example, the data is exported and available in a file

2: In the second, the information is embedded in the file

<<There exists no Face Tagging Metadata standard (ISO XXXX)>>

Within the current confines of metadata structures, surrounding names with a prefix/suffix would make extracting a name value from keyword lists at a later date, easy. For example:

holiday, Easter, ~John Smith~, party, Utah

[Regor250]

My guess about the data structure wasn't far off, and as I said a many to many relationship, tx GreyFox (that would be familiar to those with database experience). The names alone are only useful as keywords. The multiple names could easily be parsed in a coma delimited strings, eliminating duplicates and added to the ACDSee keywords if that were to be a frequently asked feature but nothing is preventing me from doing this manually right now with a series of image queries if I really wanted to. I for one did that but can be tedious if one has a long frequently changing list of people to deal with so such feature could be useful I think (call it Export People Names to Keywords under Manage People) and should optionally work for either confirmed names or both confirmed and unconfirmed.