Face tagging improvement

[mastodon]

I like to have face tagging (not recognising) like in Picasa and even better:
- drawing face tag rectangle is in Picasa more intuitive
- I can draw rectangle over the canvas (something like a bug)
- a drop down menu with names (maybe with preselectable, so I know that there where that 10 people at that event, so need onle them)
- letters should change their size to see them, or/and the nametag should move a bit from the face if face is tiny
- and? 

I am not a very experienced user, so I might just not recognised, how to use tagging in I-Match. Sorry for my bad English.

[Mario]

Define "more intuitive".

The Viewer allows annotations (face tags are just one type of annotation) to reach out of the image. This can be very useful in some cases.

There are no "tags" in IMatch. What Picasa calls tags are just one-level keywords. IMatch gives you full access to hierarchical keywords for face annotations, which allows you to automatically arrange all face-tagged images somewhere in your keyword hierarchy, e.g. under WHO|Persons|...

When you work with face annotations (or annotations at all) make sure the annotations panel is open in the Viewer. To add or change the tag (keyword) assigned to a face annotation, click the annotation and then press <F2>. The dialog shows all recently used face tags, to choose from. You can add new tags by entering them into the dialog. IMatch maps these "tags" into the keyword hierarchy you have selected for face annotations under Edit > Preferences > Metadata 2:  Annotations.

The font size of face annotations automatically adapts to the screen DPI to ensure that they are neither too large nor unreadable. Please give specifics about your screen resolution, DPI etc. Add a screen shot.

And, PLEASE: Do not mix questions with potential bug reports and comments with a feature request. Feature requests will be looked at month after they have been filed, and if you want answers or think that you have found a bug, a feature request is surely not the right place for it. It may be easier for you just do dump everything into one post, but it makes it more hard for other users and me to provide help.

I will move your feature request into General Discussions, where it belongs.

[hluxem]

I started using the face annotations some month ago and plan to replace my categories for persons with the face annotation.
I tried Picasa and Lightroom for the initial tagging and settled on LR to do the initial tagging for new pictures in the database. While it would be nice to use Imatch for face recognition, I understand that this will not be a feature of Imatch.

I do agree with the OP that the tagging in Imatch could be improved. Here are some of my thoughts regarding face annotations. I have not found a good way of adding a new annotation, when I click on annotation, I first get the choice of face detection and then a small box is placed in the middle of the image. Now I have to move and resize the box, it would be much easier to just click and drag a box around the face you want. Hitting F2 to is an extra keystroke, the text box could be highlighted after the box is created. The dialog box seems to always pop up on the same spot, while I can start typing it filters the names, but I still have to move the mouse over there to finally select the entry. Would be easier if the box pops up right next to the annotation.
I think it would be better to have the annotation outside of the box and maybe a smaller font size. I will send you a screen shot and give you more information on the dpi later.

Heiner

[Mario]

When you click on the Face annotation tag IMatch should automatically detect faces and place the annotation box around it. No need to manually place and size the face annotation box. Unless IMatch is unable to find faces.

Basically this feature is designed to display face tags added in other applications in IMatch. Unfortunately, is the face recognition technology close guarded by the big vendors and made very expensive.

I look into it from time to time, but the cheapest offer of a good and non-cloud based technology was about 15,000 US$ per year.
I don't want to use cloud-based technologies for this because these are a) also expensive and b) require IMatch to upload thumbnails of your images to the servers of the technology vendor in order to perform the recognition. Which is of course a security and privacy nightmare.

Improvement ideas for IMatch are always welcome. The face annotation displays the tag 'inside' the face region because all annotations in IMatch use a bounding box and no part of the annotation can be outside the bounding box. Snce you can just make the face rectangle a bit larger in the cases where the face tag overlaps an important part of the face, I did not want to make the annotations even more special (by displaying the tag outside the box).

This feature is available for maybe two years now. Since LR has added support face recognition support (Adobe has deeper pockets than me) it becomes more prominent and users will use it more and more. This means that I might look into face region support in IMatch soon again, adding enhancements.

There is another recent feature request for supporting the proprietary region rotation feature Adobe has added to XMP regions in LR. Instead of implementing their own standard (Adobe is a key partner of the Metadata Working Group) they apparently decided to add some extension to make it easier for them in LR.

[ubacher]

One addition which might be an intermediate step would be to at least have Imatch mark all images containing (recognizable) faces.
A show all images containing faces function. This would simplify the cataloging process.

Would be nice if we could call lightroom or the like from within Imatch to do the face recognition - a dream?

[Mario]

A show all images containing faces function. This would simplify the cataloging process.

Already implemented. IMatch puts all images with face annotations into the corresponding collection under Annotations|Regions|Face. See the IMatch help on face annotations for details.

Would be nice if we could call lightroom or the like from within Imatch to do the face recognition - a dream?

Just let LR recognize the faces. IMatch will pick them up from the metadata written by LR automatically.

[ubacher]

IMatch puts all images with face annotations into the corresponding collection

I meant that it is Imatch which scans images for faces. Imatch can already detect faces but if I understand it correctly
we need to initiate a face search manually for each image. So I was thinking of a Search all selected images for faces option.

Just let LR recognize the faces

If I want to use Lightroom only for face detection I nevertheless need to ingest all? my images into the lightroom catalog, do the face recognition, and then export
all images which were face-recognized so that Imatch will see them. I wish that this process could be simplified. ( I do not have/use lightroom myself - so have limited knowhow about it.)

[Mario]

I meant that it is Imatch which scans images for faces. Imatch can already detect faces but if I understand it correctly
we need to initiate a face search manually for each image. So I was thinking of a Search all selected images for faces option.

That feature has to be interactive. No algorithm is 100% correct and the user has to review the results of the dace detection. If the open source library I use for the core work gets better I may run it in batch and let the user review later. Not now, though.

If I want to use Lightroom only for face detection I nevertheless need to ingest all? my images into the lightroom catalog, do the face recognition, and then export
all images which were face-recognized so that Imatch will see them. I wish that this process could be simplified.

This is how it works. LR needs your images in its catalog to not only detect faces but also recognize them. It needs face data and associated tags to train its recognition algorithm.

[mastodon]

I like two tag with the mouse only, so I like to:
- resize rectangles with two sides moving despite four (like in Piacasa) - it is more convenient
- not draw over the canvas, I don't find any reason for allowing that
- have the text under the face annotation not inside, that way it is more legible and it does not cover the face, that is important in group photos where faces are small
- edit text with one right click on the face annotation text (without <F2>), like in any file commander
- select names from a drop down menu in text box, that can use with right click to the right end/lower corner of the textbox (like normal menus in windows toolbars)
- drop down menu should be preselectable (for events is better to select from 10-15 people before tagging, that would be in the menu)
- move overlapping face tags automatically
- see tooltips for the blue icons overt "Annotation layer" text on Annotation panel
- select the next image with left and right arrow on the left and right side of the image
- to check tagging with a special collection: face collection (just the faces that belongs a face tag, like in Picasa), and if I click on a face I can edit the picture in the viewer to correct tag
- I am not at my IMatch workplace, I write about font size next time (same request as hluxem)
your support for IMatch 5.5 is excellent! 

thank you

[Mario]

A general discussion post is not the place to write feature requests. It will never be read again.

[mastodon]

Sorry, but I'm not with you. You have moved this topic here. This topic suppose to be a feature request "pack".

[hluxem]

Mario,

Before submitting a feature request, would you rather have one FR with a list of different ideas or one FR for every point?

This means that I might look into face region support in IMatch soon again, adding enhancements.

That's great to hear, I like the way it works now but I do have some suggestions for workflow improvements. I usually don't like Adobe, but LR has some nice features for ease of tagging and display of annotations. For me it's not a problem to use LR for the initial face recognition and with a consistent workflow the data transfer to Imatch works well (As long as the pictures are not rotated :>). I plan on using this feature to replace my "Who" categorization, just like I use the GPS location info for the "Where" categorization.

There are no "tags" in IMatch. What Picasa calls tags are just one-level keywords. IMatch gives you full access to hierarchical keywords for face annotations, which allows you to automatically arrange all face-tagged images somewhere in your keyword hierarchy, e.g. under WHO|Persons|...

I like to create a hierarchy, but did not find a way to do this without naming the people like "Family - Fritz Mueller". I don't like to have "Family  displayed in the pictures and the text would most certainly be too long for the annotation box. I tried using a different tag in combination with data driven categories, but that will not work with multiple persons tagged in an image. Any suggestion for this or is this just not possible at the moment?


Thanks,

Heiner

[Mario]

Related ideas should go into one request so other users have it easier to follow, comment and like.

For information about hierarchical keywords and how IMatch maps the one-level simple 'face tags' you can assign to XMP regions (in your case: face regions), see the IMatch help.

Remember that the IMatch annotation system is not designed around merely tagging faces. IMatch annotations are most often used by the scientific IMatch user base, insurance specialists, librarians and other users who need to annotate images. They are most often not interested  in tagging faces, but in doing real annotations.

Face tags are just a special version of the simple IMatch rectangle annotation. I have no plans to change the entire annotation system to match the behaviors of a specific software. If you have good ideas and show me examples (remember: I may not use the same software you do, so screen shots or a short film may help) I will look into this of other others like your requests. I'm always improving IMatch. And now that Adobe has added face tags to LR, their market force will cause that this feature will be more prominently used. Before it was only the (no defunct) Google Picasa which allowed users to use face recognition. I hope Adobe sticks to the "local only, not cloud-based" face tagging for a while.

[hluxem]

I created a FR with suggestions for face tagging improvements. Please feel free to add and comment.

https://www.photools.com/community/index.php?topic=5781.0

Heiner