Reply
Thread Tools
Posts: 199 | Thanked: 144 times | Joined on Sep 2009 @ gbg.se
#11
Originally Posted by benny1967 View Post
Depends. A Map application might, for example, fully integrate with the photo browser and let you select an "overlay" (similar to google earths data sources) in which it shows thumbnails of all images taken within the area you're viewing.

Another map application might not know about EXIF or photos at all, but may be able to take latitude/longitude as command line parameters and at least show some marker at the position indicated via this mechanism. this would be good enough for locating one single image from a file manager or image browser.

a third application may be an image viewer that shows a small map in the properties window along with other exif data in textual form.

there are many possibilities, and depending on what you're up to, you'll want to choose one of them today and the other tomorrow. - this is why i would love not to be tied to the map application that comes with the N900.
Agreed. Well, a standalone map application could always display thumbnails or otherwise indicate photolocations as long as it has access to the photos and the ability to read the metadata (assuming the coordinates are stored in a universal manner, not only in the correct field but also in a universal format. Maybe that is an issue in itself). That is not bad start but for integration in the sense I am talking about, I suppose a photo browser either has to have map functionality built in or "awareness of" and "compatibility with" the map application in question, which, I suppose, is where it gets tricky. The details immensely exceed my knowledge, but I presume they involve some sort of API, preferrably standardized, for the map application. Vice versa if you would want to go from the maps application and bring up one of those overlayed images in a photobrowser. Is that where the crux of the matter lies, perhaps, a (standardized) photo-interface in between applications?

From a personal perspective I would however take integration of native applications over no integration at all. Maybe not in the long run, but surely in the short run.. Actually, with the integration pointed out by zerojay the part missing is for the maps application to be able to bring up thumbnails of images, and possibly ability to open up the photobrowser when one of those thumbnails were selected. Could we have that, please?

Last edited by nymajoak; 2009-10-06 at 15:43.
 
allnameswereout's Avatar
Posts: 3,397 | Thanked: 1,212 times | Joined on Jul 2008 @ Netherlands
#12
Seems like something (eventually) for Brainstorm?

(Although IMO waiting for final hardware/software in hands before going Brainstorm is Good Thing.)
__________________
Goosfraba! All text written by allnameswereout is public domain unless stated otherwise. Thank you for sharing your output!
 

The Following User Says Thank You to allnameswereout For This Useful Post:
benny1967's Avatar
Posts: 3,790 | Thanked: 5,718 times | Joined on Mar 2006 @ Vienna, Austria
#13
Oh, and btw:

Originally Posted by nymajoak View Post
I had never heard of IPTC before (thanks for teaching me), but apparently it contains an abundance of fields. Among other a number pertaining to contact information, including multiple email adresses. That contact information is however supposed to point to the creator of the photo, There is also a field called "keywords" which seems more aimed towards what we call tags...
Right. In fact, popular web-services extract information like title, description and tags from exactly these fields when you upload an image that contains such info. And yes, there's no field I'd know of that is exclusively meant for including "the email address of the person depicted".

Doesn't matter much, though. One could do one of three things:

a) Use tags, as you proposed. The tag wouldn't read "joe@his.net", though, but "person:mail=joe@his.net". This way it will be easy to find for applications looking for it - and will still make sense to those who only take tags as non-structured words. (flickr did this before they had proper geotagging: You could prefix tags with something like "geo..." to give latitude/longitude in an plain-text field.)

b) Find yet another field that's usually not presented to humans by photo viewing software. (Tags, titles, descriptions, dates etc. are)
It's prettier (because we never don't get to see the cryptic "person:mail=joe@his.net") and has another advantage... which i'll discuss later.

c) Use XMP. XMP was an attempt by Adobe to bring the flexibility of RDF to image metadata. I hear there are architectural issues with it, but in general people say it's expandable - which means: It has a set of pre-defined fields, but you can add your own and give them a well-defined meaning. Adding existing FOAF-attributes to XMP-tags could work, because both are RDF-based. This would save us the work of defining anything Maemo-specific. We'd just use 2 existing standards that live within the same framework. (I'm a little worried though that I don't find any examples of FOAF in XMP via google... maybe there is an issue with XMP I don't know about.)

One general issue with using mail addresses as identifiers, of course, is privacy. You tag a photo, send it to your brother, he passes it on to his friends, one of them puts it on flickr... and suddenly the mail address of the person on the image is publicly available as a tag on flickr.
This shouldn't happen.

FOAF faced the same challenge. They found a way to securely cipher the mail address so that it's still unique, but will come out as a meaningless string of letters and numbers. Only if you already have the same mail address in your contacts you can compare the two by using the same algorithm... but it's not possible to reverse the process.
So, "person:mail=joe@his.net" would be changed within the image to "person:mail=eea5dd38a317c5051438e70bdfa8f3ccfbffa 88d", and we certainly don't want to see this in any tag/title/description. This is why i think solution b) or c) would be more appropriate.

It's a cool thing, really. I remember there were discussions about solution c) on various mailing lists some years ago; don't know why I never got to see a real-world implementation. I hope it was only people being too lazy, not some real technical problem.
 

The Following User Says Thank You to benny1967 For This Useful Post:
Posts: 356 | Thanked: 231 times | Joined on Oct 2007
#14
About person on image:

What you are interested in is XMP. Exif is more for technical data and IIM (commonly known as IPTC) is considered obsolete.

In XMP you are interested in Iptc4xmpExt and tag PersonInImage this is bag type so you can put in it your own format, preferably XML of course.

About maps: I really hope N900 (and its successors) will use real GPS tags and not GSM tricks (which are evil BTW). With them things are really easy. Anyone porting digiKam to ARM?
 

The Following 2 Users Say Thank You to vvaz For This Useful Post:
Posts: 199 | Thanked: 144 times | Joined on Sep 2009 @ gbg.se
#15
On a related note, does anyone know if the photo browser in maemo 5 allows browsing photos by tags?
 

The Following User Says Thank You to nymajoak For This Useful Post:
zerojay's Avatar
Posts: 2,669 | Thanked: 2,555 times | Joined on Apr 2007 @ Halifax, Nova Scotia, Canada
#16
Originally Posted by vvaz View Post
About person on image:

What you are interested in is XMP. Exif is more for technical data and IIM (commonly known as IPTC) is considered obsolete.

In XMP you are interested in Iptc4xmpExt and tag PersonInImage this is bag type so you can put in it your own format, preferably XML of course.

About maps: I really hope N900 (and its successors) will use real GPS tags and not GSM tricks (which are evil BTW). With them things are really easy. Anyone porting digiKam to ARM?
If GPS is active and has a lock, it will tag using GPS coordinates. If not, it calculates position from proximity to GSM towers and estimates. I don't see how estimating position from GSM towers is "evil". You're going to have to explain that one to me.

Yes, you can browse by tag.
 

The Following 2 Users Say Thank You to zerojay For This Useful Post:
Posts: 199 | Thanked: 144 times | Joined on Sep 2009 @ gbg.se
#17
On a different related note, the maps integration could be extended to contacts; we know where they live, at least by address (and why shouldn't we be able to geotag our contacts when we visit them?). Isn't it the most natural thing in the world to integrate that information with a map application?
* Get a map thumbnail of their home in a contacts application
* Click the thumbnail (or a button) to bring up a map application focused at their home and/or work
* Click a button in a map application to bring up an overlay with all our contacts' names and "profile pictures" at their home and/or work
* In the map application, browse to one of your contacts, click and get to their details in the contacts application.

Oh, and about the use for photos linked to contacts; you could not only easily recall who that person in the photo was and how to reach him, in a contacts application you could directly see which photos you have associated with each person and for instance from those pictures have a natural selection of profile pics.

Is it just me or does this make perfect sense?

Last edited by nymajoak; 2009-10-07 at 04:25.
 
benny1967's Avatar
Posts: 3,790 | Thanked: 5,718 times | Joined on Mar 2006 @ Vienna, Austria
#18
Originally Posted by vvaz View Post
About person on image:

What you are interested in is XMP. Exif is more for technical data and IIM (commonly known as IPTC) is considered obsolete.

In XMP you are interested in Iptc4xmpExt and tag PersonInImage this is bag type so you can put in it your own format, preferably XML of course.
doesn't personinimage (no matter if in plain iptc or in the xmp version) refer to the name of the person only? (I only read the spec now after your post, so i don't really know for sure.)
if so, my family couldn't use it to tag photos: my father and i, we have exactly the same name. if you tag photos with this name, you can't search for me and only me.

also, in the N900 use case, we have to expect that some people enter contacts as "dad", "home", "sweetheart" and "peter's office". while the name is useless in such a situation, they may still have a unique mail address or other unique properties for these contacts.

are there examples of how to use PersonInImage with information other than the persons name anywhere? sounds interesting if it's as flexible as you say in XMP.
 
Posts: 199 | Thanked: 144 times | Joined on Sep 2009 @ gbg.se
#19
Originally Posted by benny1967 View Post
also, in the N900 use case, we have to expect that some people enter contacts as "dad", "home", "sweetheart" and "peter's office". while the name is useless in such a situation, they may still have a unique mail address or other unique properties for these contacts.
I am sure that could be solved by defining tags as pointers to the real contact post/person. Maybe not inside the metatags, but when interfacing say a photobrowser and a contacts application.
When you create a new tag in the photobrowser, give the user the possibility to connect it with a contact or, in contact application, let user choose a nickname for the contact. Let that nickname be automatically available for tagging.
 
Posts: 356 | Thanked: 231 times | Joined on Oct 2007
#20
Originally Posted by benny1967 View Post
doesn't personinimage (no matter if in plain iptc or in the xmp version) refer to the name of the person only? (I only read the spec now after your post, so i don't really know for sure.)
Specification isn't in stone I suspect in future standard will be expanded to include more details about person in image in more ordered way. Now this tag is the best thing to use IMO. Of course you can even create your own namespace and put there data but there is NULL chance any popular software will be able to process it properly.
if so, my family couldn't use it to tag photos: my father and i, we have exactly the same name. if you tag photos with this name, you can't search for me and only me.
Traditional systems used two simply methods:
John Smith I, John Smith II (don't confuse with anglo-saxon John Smith 1st, John Smith 2nd, etc.) or
John Smith (date of birth, date of death) - hard to use with living persons.
I think in this usecase the best way is to use e-mail to differentiate persons and adjust software to make e-mail addresses clickable.

also, in the N900 use case, we have to expect that some people enter contacts as "dad", "home", "sweetheart" and "peter's office". while the name is useless in such a situation, they may still have a unique mail address or other unique properties for these contacts.
Well, for "Peter's Office" you have OrganisationInImageName
But unless to use your photos as stock material I don't see any obstacles to put there not only "proper" names but nicks, relationship descriptions, etc.

are there examples of how to use PersonInImage with information other than the persons name anywhere? sounds interesting if it's as flexible as you say in XMP.
XMP is flexible but it has to be implemented. Iptc4xmpExt is quite young standard.
 
Reply

Tags
contacts, integration, maps, photos, tags


 
Forum Jump


All times are GMT. The time now is 11:41.