0 Replies Latest reply on Jun 17, 2018 7:21 PM by philc30713433

    People View: Suggested Names, Engine Progress, "Unknown Names" & Photo Stacks across multiple imports

    philc30713433 Level 1

      I've posted about the people view previously and am glad we can "rerun" the search on an image.  However, that's not what I meant in my previous comments, I meant rerun the "suggestion" engine on the unnamed folks.

       

      I run a weekly bar event. I take pictures of the people who attend and we have regulars who are weekly, monthly and yearly.  I would love to be able to import new photos.  Then, in addition to suggesting names from the previously identified images.  It would be fantastic if it would create stacks of similar faces among the people who have yet to be identified.

       

      Here are a few ideas:

      • Stack new unnamed imported images with the old unnamed images into bigger stacks. I actually can't tell if this is a feature yet or not.  I have about 6,000 images with people who I don't know their names.  It's a bar event, people come from out of town, bring a friend once, etc.  For people who later become regular, I can't tell whether Lightroom will pull all their images from previous imports and stack them with the new images I took today.  That would be fantastic if possible.  Right now I have literally hundreds of two people stacks that could easily be grouped into 25 or so images of the same person if all those pairs over time were coalesced together.
      • Add a progress bar or "analyzing" indicator when the recognition engine is running. Currently it cannot be determined if the recognition engine is spinning on the unnamed people trying to make suggestions.  Let's say I have 100 pictures and I know about 80% of the people.  There should be a way to "stop" the engine or indicate "I cannot name this person at present".  This would allow the recognition engine to stop working until new images were available and save RAM and CPU resources.
      • Be more explicit about the functionality (and what isn't functional). A lot of folks aren't happy with People view, not because we don't like the service. It's just hard to be satisfied when you don't know what are the features.  It's a CSI: Miami / 24 tv show expectation problem... Customers see these fake impressions of facial recognition and expect the moon.  If Adobe were more explicit and say, "This is where facial recognition is strong, we are working on improving these features, and ya know what X, Y, and Z are just hard and we're not currently improving that aspect."  This way we can not complain about things that we know you have no intention of improving.
      • Editable tolerance for recognition. I photograph a lot of bearded white men. Brown beards, gray beards, stubble to ZZ Top... and periodically people even shave.  I would love to be able to say to Lightroom, "if you were being less strict about matching, re-match the remaining unnamed photos".  Even more amazing would be to say, "Hey this guy is wearing a hat, or "that's a wig don't use information about the eyeline". I figure the later part is a pipe dream.  However I do think it would be amazing to see "Hey be more aggressive on matching because there were only 10 distinct people at this shoot, and then say "Hey this is a stadium shot, there are 200 distinct people and I expect you know 60% of them."
      • Social Network data to improve matching. This is where facial recognition gets a little intrusive, but it is a fact that people run in cliques. The suggested names should utilize the past social clustering information to inform the future suggestions. If Adam, Eve, Cain, and Abel are almost always photographed together, please rank Cain as a more likely candidate when Adam and Eve are present in the shot.  Yes, the hair or something may be different about the face. However if there's a 80% match to Zeus and a 60% match to Cain, maybe Cain should get a "social network" boost by the others identified in the shot.
      • Multi-word tags. It would be great if the tag suggestions didn't restart at the beginning of the word following a space. Currently if you type "Santa Cl", it will start to suggest "Claus Gutenberg" rather than completing the keyword "Santa Clause".

       

      Thanks for listening.

       

      I love the service and am happy to give suggestions for improvement. It would be great to get more active feedback or comments from developers so we don't go another 3 years waiting for an update that doesn't really address a lot of peoples' prior suggestions and dissatisfaction.  Adobe, we love your products, you know you have a loyal base, improving communication will help us appreciate you more instead of wondering what's going on in the black box and then being frustrated when years worth of suggestions go unheeded.