| DO NOT USE face tagging as your only means of identifying people in a photo. It's NOT a standard and no matter what system you use, it's unlikely to be supported in the future. Now read on to find out why. |
Tagging - I'll start with a cautionary note about the use of the word "tag" which has many meanings. To "tag" a photo generally means the addition of keywords to the image (it also has a host of other meanings). What we are talking about here is Facial Tagging or Face Tags, identifying a face in a photo and attaching information (i.e. name) to it.
Before we get into a discussion of facial tagging we first have to talk about standards, since that is the crux of what this discussion is all about. Will what you do to a photo be viewable/retrievable 50 years from now?
Archival Standard: A face tag, in any form, IS NOT an archival standard. My definition of an archival standard is one that meets at least three criteria:
- That it be an open standard (non-proprietary), one with its definitions clearly defined, available to all and supported by an active organization (a group in control of the standard such as IPTC supported by the International Press Telecommunications Council).
- It has to be widely adopted by software developers. It takes years, even decades for any system to develop into a standard. The digital world is full of "better ideas" that have fallen by the wayside.
- It has to be supported by the current biggest player on the block, which in the case of digital photos is Adobe Inc. That's just a reality of the way the world works. It's very hard to call something a digital photo standard if Adobe doesn't fully support it.
You'll note that I do not include "adopted by archival institutions" - since they have more rigid guidelines and different goals than a genealogist. For me, JPEG is an archival standard simply based on the fact that is an open standard adopted by virtually every piece of image software on the planet. JPEG can incorporate IPTC/XMP metadata within the image file without any change in the quality of the digital image (same as TIFF - many other image file types cannot incorporate metadata). JPEG images I created in the 1990s will be viewable in the 2090s. But an archival institution will usually pick JPEG2000 over JPEG due to its slightly superior image quality (we're talking the best quality settings for each), whereas I think that a high quality JPEG will serve 95% of the time. See photo filetypes. Those interested in true archival standards should check the US Library of Congress website.
Face Tagging
There is no standard for face tagging, different software, different on-line services use different systems. The Metadata Working Group attempted the development of a standard to define face tagging, but that seems to have been abandoned. If you do face tagging in one system (i.e. Microsoft Photo Gallery), there is a very high risk that it won't be supported by other software. There is also a risk that it may not even be supported in the future by the company that developed it. Think of software such as Picasa that was abandoned by Google in 2016.
Part of the reason is that we're in early days (2020) of these systems, software developers are coming up with lots of "better ideas" when it comes to face tagging. New systems using AI (machine learning) facial recognition are already making most face tagging systems obsolete.
Other than fun, you should question why you want to use face tagging, what advantage is it giving you over conventional labeling when it comes to genealogy. What you're trying to do is to properly label the photo and ensure that the labeling will be viewable/retrievable (by anyone) 50 years from now. Since face tags don't meet the criteria for an archival standard, odds are very high that any face tagging you do today will not be viewable even 20 years from now, much less 50.
AI will make face tags obsolete. How can I make such a definitive statement? It's because it's already happening. There is facial recognition software based on AI that, with a single face identification can pick that face out in other photos. So, why bother face tagging when software is going to be able to do that automatically for you. However, all that is based on initial identification of people in photos, and that still points to properly adding proper archival labeling to each photo which is the IPTC/XMP standard.
Down the road, AI may well be able to look at your IPTC/XMP metadata description of the photo and with just a little bit of training be able to highlight those faces (do its own face tagging based on your IPTC/XMP description) in all your photos. It should also be able to identify that face where you've been unable to identify it and also point out errors in your IPTC/XMP description (where you identified Great Aunt Millie as Great Aunt Lyla). It just depends on how "smart" AI actually gets (I'm guessing "pretty smart" :-) and how it is implemented in photo and genealogy software.
I also worry about the privacy issues surrounding face tagging. When used on-line, that information becomes available to everyone, including AI software. This is already AI facial recognition software in use that uses a database of social media postings - if you've been identified in an on-line facial tag (any on-line system that uses facial tags) then odds are that you're now in an AI database. Government control (to ensure the protection personal privacy) is lagging far behind the technology.
Personally I don't do face tagging and don't plan on using it. Bottom line is that I still advocate doing boring old conventional digital labeling using IPTC/XMP if you want to be 100% sure of archival longevity. The pinnacle of that is to do visible labeling which isn't dependent on any software/system or standard other than the photo filetype (i.e. JPEG, TIFF).
|
