Object Recognition - Untagged


I love this video! What you are looking at is a proof of concept for a gesture and object recognition context-aware user interface by Beverly Harrison and her team at Intel. It is a simple plywood surface and some Legos. None of the items are tagged, they are ordinary, household items. What Beverly and her team have done is create a smart system that tracks objects on a surface. At the beginning, you see her trace her finger on the table and train track fills in. When she moves the train around the track to the station, if you look closely at the surface, you can see little people exit the train.

By using these robust algorithms the team can author any behavior for any object and dynamically re-assign its role. For example, the people could run, walk or move in any pre-coded direction. The dragon that is part of the scenario can be designed to blow fire, bubbles or smoke.

The surface can be large or small. Larger surfaces may require the camera to be further away which in turn degrades the accuracy. The distance is determined by the needed level of accuracy.

What would you do with this technology? Some of their suggestions were tracking items on a workbench or machines in a manufacturing shop. Personally, I love the opportunity for open play with Legos!

Comments