What meta-data do we have for the average digital picture we take:
- MINIMAL:
- a filename, typically autogenerated by the camera (e.g. “DSC0009″) or chosen at the moment of import (e.g. “Trip to Portugal 001″ or “Aug2005_001″)
- a filedate, which probably correponds to the date the picture landed on the hard disk
- EXIF information: date of image capture, camera brand and model, aperture, … (maybe in the future also geo-location from a built-in GPS)
- ADDED BY HUMAN HAND
- a title and a description: in free text
- an group/set/album name: typically less than 10 words
- tags or labels: the ideal search criteria, typically added by the owner
- geo coordinates: the new craze on Pixagogo (who then also add the city name as a tag), so the pictures can be mapped on Google Maps
If the human-added metadata is missing, there is hardly a way to find the picture through Google Images or Flickr. What if there could be an software that analyzes a picture and automatically adds relevant metadata to a picture?
Munjal Shah, onetime cofounder of the auction services firm Andale, finally let slip on his new blog what he’s been working on since leaving last year (…) In other words, his startup, tentatively named Ojos (Spanish for “eyes”), is creating a new way to search and organize photos.
(…) he revealed the key technologies behind Ojos: face and text recognition. (…) The other key: You can assign tags, or keywords, to one photo and the service will automatically append that tag to other photos of the same people.
blogs.businessweek.com
I wonder if it also could be used to recognize familiar archetypes/icons like: a house, an sunset, an iPod, a Ferrari…
On his own blog, Shah writes:
I think Flickr’s tag based system is just super (in fact I love it), but I wanted all of my photos on there, I wanted them all tagged, and I didn’t want to spend hundreds of hours doing it. So being the lazy engineers that we are, we thought maybe we can at least auto-tag some of the faces and names.
on munjal.typepad.com
Ho John Lee states on his blog that the technology should be offered as a web service, not as yet another photo storage site. He has a point, and I can see also it working in a technology licensing model: let Flickr or Pixagogo run it locally and let them pay per million pictures treated. Anyway, it will be interesting to see where this company goes.
(via John)
Technorati: AI – images – startup – technology
Related posts:
- Folksonomy and google bombs Folksonomy or social tagging A folksonomy is a system that...
- Web 2.0 mememap overview After I saw Tim O’Reilly’s Web 2.0 mememap (via readwriteweb.com)...
- What Google Agenda currently misses I am using Google Agenda as the central repository for...
To answer your question: Object recognition for generic objects like cars, etc is not within the limit of technology today but you can find similar photos. So you can highlight and tag a red BMW Z3 (side view) and we can find other red Z3s based upon our similarity technology(side view) but we can’t say for certain that it is a Z3. Of course then you could select 24 of 30 we return an quickly tag all of them as a Z3. – Munjal shah
http://cornell.elliottback.com/archives/2005/03/16/want-to-join-a-startup/
This post has been removed by a blog administrator.
This post has been removed by a blog administrator.