Fast and Low-Power Mobile Face Recognition
Combining all three of the major themes of the research projects that we saw - facial recognition, mobile, and cloud-computing - this next project firstly aims to create an effective real-time facial recognition system that will work on low-power mobile platforms such as netbooks. The demo we saw showed it recognising and tracking the demonstrator's face at a rate of about 5 fps.
The next level of the project adds in the ability to switch the calculations from the mobile device to a cloud-based environment. With a single command, the system was able to seamlessly switch from local processing to cloud processing and boost tracking performance to about 8 fps.
Quite how useful such a system could ever be is anyone's guess. Quite aside from the fact that there would need to be some standardised way of doing this for it to take off, we're not even sure what mobile devices require real-time facial recognition. Any suggestions?
OASIS: Smart Computing Islands on Everyday Surfaces
The final project we looked at was showing a system that allowed you to manipulate virtual items projected onto any surface. By using a camera that picks up RBG data, for tracking objects, and depth data, for distinguishing between hands and other objects, the system can identify small items like a fake plastic steak and an orange (well that's what we saw) as well as track your fingertip as you move it across the table's surface.
Once an object is recognised, a context aware menu will be projected next to it. You can then use your finger to select items in the projected menu and proceed to, say, open a recipe page related to the item, or watch a video about it. When you're done interacting with an object you can store it in a virtual draw ready for reusing later.
Watching the demo, this is the sort of tech that really has a wow factor, even though many of the techniques that make up the overall demo have been around for a while now. However as with many of these projects and indeed with science generally, it's not necessarily about great leaps forward. Sometimes it can take years to fine tune a system/theory/technique to get it to a stage where it's ready for public consumption. Will we see interactive desks with projected objects or ray-traced games playable on our netbooks anytime soon? Almost certainly not, indeed we may never see some of them as they're deemed impractical but it's certainly nice to see someone exploring new avenues and pushing forward frontiers.