Revamped Google Picasa site identifies photo faces
http://news.cnet.com/8301-13580_3-10026577-39.html
(CNET News, September 2, 2008)
Google has announced a new way to automatically tag photos containing human faces. This feature is called “name tag” and suggests proper tags for a picture once you tell it who is who.
Those who have used the iTunes after organizing their songs with tags may understand how tagging makes it easier to find and play songs. In spite of the convenience that tagging delivers, what makes users reluctant to put tags on their contents in various forms was the tedious nature of tagging work. In this sense, the Google’s new feature could be one of breakthroughs to turn untagged contents into tagged ones.
According to the article, once you’ve started identifying people in your pictures, you can tag about 200 persons in more than 100 pictures just in mere 15 minutes. Imagine how convenient it will be if you can retrieve a proper set of pictures from your huge picture storage by just putting in the name of the person you are looking for.
Despite some erroneous suggestions are made by this feature from time to time, I think that this attempt shows a new possibility how people can organize contents by tags more easily in the future.
Relevant lectures:
8. Classification
15. Personal Information Management
18. Metadata for Multimedia
27. Multimedia IR
Kentaro Suzuki Said,
September 7, 2008 @ 5:07 pm
This article sound very interesting to me. Also, it reminds me of Google’s other service– Google Street View.
Google has been developing a method to detect human faces, which is applied to Google Street View to blur human faces taken by Google Cars. Therefore, it is not so difficult for Google to apply this methodology to Picasa to detect human face’s picture.
Of course, it is more difficult to distinguish a certain human face from others’ than to distinguish human faces from other objects. However, if I use IR words to assess IR systems, it is not needed to mark the perfect point in term of “Precision” and “Recall”. Some failures in recall and precision don’t make so much a big problem. People can easily delete pitures of faces which are different from the face they want to search and put a name (in terms of “Precision”), and even if there are some pictures which can’t be detected by Google, it will not make big problems (in term of “Recall”).
However, in the case of Google Street View, some failures to detect and blur human faces may result in fatal privacy problems. Therefore, Google Street View requires a facial detection system which has a very high mark in term of “Recall”.
I would like to know how high mark Google’s method for the face detection has, in term of “Recall”.