Start a new topic
Planned

People Tagging

Some suggestions for people tagging:


Show a bigger image so it is a bit easier to tell, some people, specially young kids with siblings are tough to tell :)


Ability to rename people you tagged in case you made a mistake while entering the name (and to merge with existing names)


While entering the name - auto-complete to prevent typos and differences in how a name was entered


Will it do automatic facial recognition?  It didn't seem like it was doing it at all.


27 people like this idea

In response to: 

"Your (3) suggestions are very welcome. The face size may be dependent on the device you are using to a certain level. Its should be doing automatic facial recognition by grouping people."


The face size is primarily dependent on the little circle that you embed the face in. That circle should be bigger. There is certainly a fair amount of white space around that circle. Even better would be to allow us to see the whole picture so that we can see context, clothes, etc. to identify the person. It would also be helpful to allow us to scroll through all the images in the group. For instance, you may show a face and say it had 7 related images. If we can't tell for certain who it is in the little round image you show us then it would be nice to scroll through the other 6 as that might make us more certain. 


1 person likes this

I have to agree with the most recent comments.  I still love the idea of Monument but the facial recognition was a HUGE selling point and it does not nearly meet expectations.  Right now it recognizes a face, but it certainly is way below expectations in recognizing people - it is without exemption that I have to constantly merge faces and I'm only uploading a few pictures at a time.  I don't dare move the thousands of pictures I have from before monument because I don't want to spend a month tagging people.  I really think you have to improve how the system recognizes people if you expect to really "sell" that feature, I would have expected something better than what picasa was able to do, and right now I'm sad to say it is a longs way from that.


Another quick point - considering that we have to identify people in almost every picture we upload because monument isn't able to do so automatically (well and for now only hopefully) at least give us an auto-complete feature when we are entering names.  


I'm using this on iOS for your reference.  Thanks for the hard work you have put into this device and for all the work you are still doing to make it better.


1 person likes this
If you're going to keep saying merge improves accuracy then prove it. If anything it's gotten worse. I can't tell you how many random things get picked up as people. Why is my iPhone about a billion times better at recognizing people? This device has very few jobs and it's not good at many of them.

1 person likes this
I made a similar comment a month or so ago. Without real, learning, face recognition, this product is DOA. It needs to be as good or better than Facebook tagging or iPhoto. I have no idea if Monument hardware is robust enough to run the required code. I hope so.

1 person likes this
Every time someone brings up this problem, your solution is to simply join images with the same face. That might be okay, if Monument LEARNED, but it doesn't. In my case I have more faces with one single image than any other category, but, of those single image faces, hundreds, maybe thousands, are of the same person. So, your solution of joining them would take days, weeks or months. That is just not satisfactory! The ONLY thing your algorithm seems to do is recognize when a FACE is in a picture, and it seldom recognizes the same face. Also, I have found that in a group of pictures that it found the same face, OFTEN the faces are NOT the same. So, even after grouping, it can turn out that Monument made mistakes and grouped different faces. In this case, not only do we have to manually join faces, we then have to manually separate mistakes! This is simply ridiculous.

Jim

 


1 person likes this

@John: If after you select them after hitting edit, and then subsequently hit merge, they will be merged. Then you can name the merged face, and look into it in the named faces section, and remove faces that may not belong in there. Once you do that, the accuracy increases.

See inserted image showing faces that are of the same persons but not recognized as such. Different colors identify different people. App Version 1.5.1 Monument Version 1.2.2

image

IMG6793.PNG
(2.25 MB)

I am having the same problem. There hasn't been a larger group of photos for any one person than 10. I have 60,000+ photos. One way to learn more from your users is to track in diagnostics all the different merges performed on any person. Then make it possible to upload the diagnostics to you as well as the face you showed as the selection for each group. That should give you data so that you can improve your facial recognition. 


What you have done is pretty remarkable. I appreciate that I have the ability to post this complaint because your software is making an effort to recognize faces already. I think it's just going to require constant improvement until you are able to recognize the same face in 100's and then 1000's of pictures automatically. 

I use a Mac, with the Photo app I have already tagged many faces, it would be great if those tagging could be imported in Monument, it would save us a lot of time!

@Yves; just to repeat, we do NOT have access to your phonebook. we pull up the keyboard built into the OS, and if the OS supports it, the keyboard provided by the OS will fill in the suggestions provided by the OS. We have nothing to do it. It behaves the very same exact way for all Apps.


  • You having control and thus acces to my phonebook, is not something I want from photo software. We don't have access to your phone book, your OS may. Don't like it, change your OS/Phone.
  • You don't need it for this kind of software. Our software does not do this, and we don't need it, and we have not implemented it.
  • I don't want to link pictures to contacts. Not sure where this link thing come from. We said no such thing and we do no such thing.
  • that invades other peoples privacy. something I don't want to do. Once again, we don't have access to your phone book.

We hope this provides sufficient if not overabundant clarification that we are not doing anything you claim we are doing.


You having control and thus acces to my phonebook, is not something I want from photo software. 

You don't need it for this kind of software.

I don't want to link pictures to contacts. 

that invades other peoples privacy. something I don't want to do.



y

Would this not simply require that the Monument app ask for permission to access the address book?

I really really really think it is important to have a button to use to identify inanimate objects that are NOT faces, especially if your algorithm leans face recognition by the way it matches faces as they accumulate.

 

@Yves: that appears as suggestions for names matching a name from your phone book. The OS does that independent of our software. So, if I have your name in my phonebook, and start typing yv, yves is suggested by the OS in some OSs. We have no control over it.

>This exist where the OS itself (for example Android 6, 7) provides the names from your, for >example, phone book. 
I don't like the link between pictures and people in my contact list.


@Clay, please refer to below:


  • Auto complete from existing names would be very helpful. This exist where the OS itself (for example Android 6, 7) provides the names from your, for example, phone book. But does not exist in other cases. Currently, we rely on the OS to provide this support.
  • Renaming existing names would be very helpful. Already there, and has been since last week's release.
  • Viewing the contents of a facial recognition set where there is no face (false positive) would  also be very helpful. Interesting.


1 person likes this
Login or Signup to post a comment