On Fri, Jul 8, 2011 at 9:13 AM, Luciano Resende <[email protected]>wrote:
> On Sat, Jun 25, 2011 at 1:37 PM, Subash Chaturanga <[email protected]> > wrote: > > Hi devs, > > > > I thought of a way of providing UI functionality to this and updated the > > JIRA[1] . > > > > There, the UI panel which gives users to train their friend's images. And > I > > have created a seperate html called face.html [2] which is directed from > a > > link through gallery [3] . WDYT? > > > > [1] - https://issues.apache.org/jira/browse/PHOTARK-73 > > [2] - > > > https://issues.apache.org/jira/secure/attachment/12483826/Screenshot-2.png > > [3] - > > > https://issues.apache.org/jira/secure/attachment/12483827/Screenshot-1.png > > > > Great progress, I was wondering what do you expect to be the end user > experience when using the face recognition service, or do you have > some documentation on what users should do to train the system ? What > I had in mind was something like, when navigating a album, I could > select a "square" in the picture and type the name of the "friend" or > "person" that the face represents, very similar to what you can do in > facebook, or other social applications. Did you have something > different in mind as for the user experience ? > Hi Luciano, As I feel you are talking about training the face index before recognizing, because in recognizing, we just provide an image and give user names and check whether there are in the picture. So before recognizing, we need to train the face.com index by our target's tagged images. For Facebook Friend Finder, we can have a straight forward train by only providing the userid of my FB friend . Then this system will get their tagged images and train them. In general scenario, to train the index, we have to provide image(s) and ask the system to detect them and it will return tag id(s) and then we can save those tags under a username with the system provided private name space. So your scenario counts when it comes to general face recognition under a private name space. There, the typical scenario is, train the users by your albums images. Hence I thought of providing a separate place to do index training stuff, where you can give image location, or url, and train particular users with their private namespace user names(as shown in previous screen shots). So in your opinion, you want to train the friends by only specifically providing their surrounded faces in those local album images other than the images itself. It seems pretty nice and cool.But I am currently not aware of the technology which should use for that. Shall I look in to that and do some research on it to find out the possibility of doing that. WDYT ? > > Thanks, > > -- > Luciano Resende > http://people.apache.org/~lresende > http://twitter.com/lresende1975 > http://lresende.blogspot.com/ > -- Subash Chaturanga Department of Computer Science & Engineering University of Moratuwa Sri Lanka Blog - http://subashsdm.blogspot.com/ Twitter - http://twitter.com/subash89
