I recently experimented with Picasa's (Google) new face recognition software. For me, it's 90% sensitive and specific.
The face-recognition feature is called “name tags” and it automatically group photos together based on the people in them.
Here's how I tested it.
All the photos used in my blog are uploaded to Blogger, which stores them on Picasa (since Google owns both sites). I clicked the Add Name Tags button on the right side of the screen and the Google server began sorting my 70 megabytes of photos. It clustered shots together based on its best guess of the people in them. It identified three major people in my photo collection - me, my wife, and my daughter. It then asked me to tag these three folks and then clustered the shots by name into easy to navigate slide shows of similar faces. A few photos such as the shot of Ney player Rifat Varol from Istanbul did not cluster and I clicked the skip button to ignore these outlying photos. A few photos of me in profile did not sort automatically and I labeled them manually.
How does it work? The software performs a few basic measurements such as the relative distance between a person's eyes and nose/nose and ears.
With the new Picasa site, you can also choose to make your public albums searchable. This lets people locate your public photos by searching Picasa Web Albums or other Google services such as Image Search. Or you can keep you albums private by clicking on the Unlisted radio button.
I used a Mac with Firefox to test all these features. I'm told that the site also supports a Windows-only upgrade to Google’s Picasa photo management software, which offers photo-editing capabilities, but I did not test this.
Automated face recognition on Google's photo management site. That's cool!
No comments:
Post a Comment