Netvibes people feed!
Thanks to the Netvibes Universal Widget API, you can now show photos from Polar Rose on your home page (Netvibes/iGoogle), desktop (Apple dashboard/Vista sidebar) or even your website or blog.
You can personalize the widget by setting the name of the person, and show either a thumbnail of the photograph or a mugshot of the face. If Polar Rose doesn’t currently have any photos of this person, you can easily tag some photos with our plugin and they will automatically appear in the widget!
July 15th, 2008 at 12:54
Is there still the plan to implement auto-facerecognition?
Never really heard from it since the very first Video making a 3D model from a Photo.
To much other work to do?
Does not yet give the expected results because the engine still needs some enhancement?
Hardware to weak for crawling, analysing and storing everything?
never the less: great job you do :)
July 19th, 2008 at 19:23
Hi Cougarten,
Face matching is still on our top list. It is coming under different forms: clustering pictures (stacks in our search results), guessing name of unknown faces (name suggestion, automated indexing), finding similar faces…
While it’s relatively easy to have something working on a few thousand images, but scaling it to millions or billions of images take a little time.
That’s what kept us busy, building a vision technology and an infrastructure that scales. But we are getting there ;-).
July 21st, 2008 at 13:57
thank you :)
July 23rd, 2008 at 07:57
dsf
July 27th, 2008 at 16:32
Since Amazon introduced S3, infrastructure is no longer an issue and you can focus on your core technology.
July 29th, 2008 at 13:52
Thomas: If it were only so simple. The Amazon web services are fantastic and we use them with a lot of success for many tasks. Unfortunately there is no auto-scaling magic in them. You still need to build your application logic, data storage strategy and infrastructure on top of it. In a scalable way.
The Storage Service that you mention is probably the easiest to use. We store images in S3 and serve them out to you through a CDN. That is one of the most simple components we use though. Other things are much more difficult to scale and require some really deep thinking and engineering efforts.