People Hopper is a new application for Orkut that lets you take your profile image and 'morph' it into your friend's–using publicly available images from other Orkut users along the way. It is a fun app that also allows you to make new connections–since the images that show in your Hopper all come from people who have chosen to make their profiles public, you can click through to learn more about them and reach out to them. You can also morph one friend into another. No computer graphics are used to generate the path.
How can I access the People Hopper gadget?
To be able to use People Hopper, one needs to have an Orkut account. If you don't already have one, create a new account here. Then, you can add the gadget by clicking on "add apps" under the "more" menu, or simply click here to:
How to use?The gadget starts with loading and displaying your friends images on the page. It shows two target boxes displaying "drag a friend here". One can drag a friend's picture in any of these two targets. Once there is an image in each of these boxes, the gadget returns the best path which transitions one face into another smoothly. Each face along the path comes from a Orkut user's public profile image. One can click on any of these path images and go to the profile of that person.
How does it work?To generate a 'morph', People Hopper uses image matching technology. First, faces are automatically detected in public profile images and normalized for improved contrast and size. Then, for each image, we find other publicly available profile images that are similar to one another. Then, when you pick a friend you want to be your end match, we just hop along similar public images, step-by-step, until the connection is made! You can think of it as an image graph between friends, similar to the friendship graph already on Orkut. We don't use any face-specific features or metrics during this process. It is just a simple matching to find similar pictures.
Will the path between a pair of images remain the same forever?No. We plan to update our database of user profiles every few days which will likely result in a new path between the same pair of input images. So, visit frequently!
Do you do face recognition in this work?No. The similarity matching does not use any face recognition technology. In other words, it can say how similar two faces are but not who they are.
Why is the gadget not showing all my friends?Before loading a friend's profile image, the gadget tries to make sure there is a person with visible face in that image. Only those friends for which such a check passes are shown in the gadget. Since it is an automatic process, the algorithm may make mistakes. We make efforts to continually improve the detection algorithms.
Why is it taking too much time in returning the paths?The time taken in returning the best path depends on several factors. Typical path computations may take anywhere from one to twenty seconds depending on how far two profile images are in the space of all Orkut images that are public. Other issues like network speed also affect the performance. If you see it running for more than a minute, perhaps it is time to hit 'reload'.
Why is the transition in my path not that smooth?The quality of the path depends on two main factors: the appearance of faces in profile images of Orkut users and the similarity measure used for image matching. Since we cannot control the appearance of profile images, it may happen that for a particular image there exists no visually similar image in the database. Developing better similarity measures is a topic of continuing research.
Are my generated paths stored once I leave the app?Currently they are discarded when you leave the app. If you want, you can store them under My Updates by clicking on the button 'Show Path in My Updates'.
What if I do not want my face to appear in any path?You can opt-out by following the instructions given at Opt-Out Settings.
It would be nice if... FeedbackWe are currently experimenting with technical and social aspects of hopping across a large set of Orkut users. We look forward to hearing from you about how we can improve it. Please tell us your views.