Phone-to-Phone Communication for Adaptive Image Classificati

We present a novel technique for adapting local image classifiers that are applied for object recognition on mobile phones through ad-hoc network communication between the devices. By continuously accumulating and exchanging collected user feedback among devices that are located within signal range, we show that our approach improves the overall classification rate and adapts to dynamic changes quickly. This technique is applied in the context of our PhoneGuide system -- a mobile phone based museum guidance framework that combines pervasive tracking and local object recognition for identifying a large number of objects in uncontrolled museum environments. We explain a technique that distributes the user feedback information during runtime through ad-hoc network connections between local devices. By doing so, we enforce cooperative classification improvements during the actual stay of the visitors. The general functionality of our technique has been tested with a small number of real devices in a museum. For proving its scalability, however, we have developed a simulator that evaluates our method for many hundred devices under several conditions. The simulation parameters have all been gathered in a museum, and are therefore realistic. We will show that ad-hoc phone-to-phone synchronization not only leads to higher overall classification rates, but also to quicker adaptations to dynamic changes during runtime.

Category:

Science & Technology

Tags:

• phoneguide
• phone-to-phone
• p2p
• mobile
• phone
• museum
• guide
• image
• recognition

[arBUW]

What do you mean by tools?

The whole simulation application was implemented in Java. It runs in combination with the WTK mobile phone emulator.

[fxp007]

What tools do you use?