A face tracker which starts learning online and does not use any offline data as in haar cascade classifiers. This is invariant to rotation , scaling and occlusion. The face is re-tracked if it goes out of view at the instant it comes into view. The tracker learns the face better with time and after some time, it can detect and track the face from sideways or even at a much occluded view. This face tracker is unique in the sense that it tracks only Â the specific face and it distinguishes other faces. This can be used in applications which require verification or applications where only the master/owner/commander or a registered user is to be tracked. Use full as a tracker for personal robots which is more attached to it’s master.