when: from 20th september 2011 to 20th november 2011
what: Democratizing Technology is an exhibit sponsored by Ford and aimed at making new car embedded intelligent systems easily understandable, through a series of hands-on experiences.
I was responsible to create two interactive installations that could educate kids (and grown ups) in a playful way.
The first one tries to show in the simplest possible way how a computer can parse vocal commands.
Visitors can visualize the sound spectrum of different commands, pass their data buffer to the sound card and hear their sound.
Then they are asked to repeat the word they heard and, while their voice gets recorded, they can visualize in real time the “shape” of their voice.
Visitors are asked to visually compare the “shapes” of the two different sounds and look for similarities and differences; finally the software explains how specialized techniques can be used to efficiently automate this kind of comparison.
The second one gives an explanation of how traffic sign recognition systems work.
Visitors are asked to imagine and draw their own traffic sign on a blank sticker and to place their creation under a camera.
The software takes a snapshot of the drawing, applies a set of feature extraction techniques and visualizes the interest points it finds (ie: edges, corners, gradients and their relative distribution inside the image). As Visitors begin to understand how these detected features “describe” the image they have just created, they are asked to give a name to the “traffic sign” they created and to challenge the computer vision software: they are asked to move, rotate, partially hide their sticker, while the software continues to track it and recognize it.
Finally, visitors are asked to choose one of the available backgrounds depicting different scenarios (urban, seaside, desert, etc) and to glue their sticker on it: just like in real world scenarios, the software still finds and recognizes the drawing, even between all the other elements in the picture.
what else: given the name of the exhibit, it was a natural choice to use only open source libraries; if somebody is curious to look under the hood, the traffic sign recognition software is licensed GPL and its source code is here.