I am currently working on augmented reality face recognition UX and I find it rather annoying that information is taking so much space if the number of matches is high when targets are close to each other. It is literally obstructing the view.

I would like to pick you brain to see if anyone would like to suggest a more user-friendly approach of presenting data to the user.

Here is an example of current solution:

enter image description here

enter image description here

  • 1
    Have you played WatchDogs? I'm sure they had to tackle a similar issue with the face recognition. Been a while since I played so I can't exactly recall what they did to solve it. Mar 6, 2018 at 14:59
  • I will definitely spend some time playing WatchDogs, but I think the difference between games in life is that in life you need to be way more alert of your surroundings due to inability to revive on game over. Mar 6, 2018 at 16:26
  • Are first names needed? Could you just show M.Samuels instead? Might help save some space. BTW, I didn't know Mrs Jones window-shops at Claire's. : ) Mar 16, 2018 at 8:52
  • I would also like to ask for what purpose/context is this in? is it for looking for criminals? I think the solution will depends heavily on the context. Mar 16, 2018 at 9:46
  • @RobE haha! Mrs Jones is probably shocked that Claire's still exists. As per your comment, yes. The name as actually taken from the phone book. But I will speak to SH, we could maybe loose surnames. I mean, you know who Mark is if you know him.. I guess. Thank you for your suggestion! Mar 16, 2018 at 10:05

3 Answers 3


Not only the names obstruct the targets, but the pointers also obstructs the targets' faces.

Multiplayer video games like PUBG stack the players' names at the bottom of the view and move on the X axis below the players. It's a good solution, especially if you have to identify dozens of targets in one screen.

Now you may ask yourself what is the user looking for? I doubt the user will attempt to read the name of every moving target. The user will likely look for specific characteristics through a search, and only the targets exhibiting these characteristics will be highlighted and named. So the names may be irrelevant until the user makes an inquiry, then only positive and negative identification may be relevant.

  • The way it currently works is that it compares the database of faces, picks the ones you marked against your phone book and then when finds the ones that you "know" it pops the names next to them and highlights the ones that matched in green and the ones that did not in red. I guess the red highlighting is definitely going as it is useless, but I thought the green one would help attract attention. Mar 16, 2018 at 10:17
  • If the people are in my address book and I know their names, why do I need them repeated on an AR display? AR should provide info that is unknown to me or difficult to recall. Mar 16, 2018 at 22:51
  • I mentioned this in other comments - because you don't see/recognize/remember faces very well. Mar 19, 2018 at 14:23

Given what's been discussed in the comments, I would suggest you take a good look at the info you want to display (names) and try to cut it down to as little as possible only to the bare bones of what's needed.

E.g. If first names will do then just use first names.

Since it's eyewear I would suggest to explore a hover/look-over 'Gaze' feature. Where the information is shown on targets that are being looked at. This is a natural way, as humans, to find out pieces of information.

Here is a Hololens tuitorial demonstrating this interaction.

enter image description here


Display on demand

If there's a lot of useful information, the user will want access to it. This will only be a problem if the info is persistent or appearing when it's not needed.

Provide users with a series of voice and device button controls to invoke this data only when it's desired. This allows the user to have an unobstructed normal view with quick access to all the AR power your app provides.

Voice is convenient (when it's reliable) but users don't always want to announce what they're doing with their new stealth smart glasses (Vuzix Blade, anyone?). Providing a simple interaction model for hardware buttons is a must in such scenarios.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.