One-Shot-Learning Gesture Challenge
Source: Alex Armstrong

Microsoft has become a sponsor of a contest to devise an improved gesture algorithm that will analyze a series of Kinect video streams. The prize fund now stands at $40,000 USD.

The ChaLearn Gesture Challenge has opened on Kaggle, a website that hosts predictive modeling competitions for cash prizes.



gesturebig



The challenge is to discover whether machines are capable of "one-shot-learning" i.e. recognizing patterns such as hand gestures after seeing a single example.

Applications for such pattern recognition include man-machine communication, translating sign languages for the deaf to hearing people, and computer gaming. While the primary focus is hand gestures, facial expressions and whole body motion may enter into account and the contest uses Kinect video clips as its data.

In the Kaggle round of the contest, which continues until April 10, 2012,    judging is done algorithmically by calculating and comparing the actual and predicted outcomes.

Two further rounds of the competition with "live qualitative evaluation" are scheduled to take place at two pattern recognition conferences: CVRP 2012, taking place at Providence, Rhode Island, June 18-20, 2012 and ICPR 2012 in Japan, November 11-15, 2012.

   

gesture



The competition is open to any "professional or enthusiast in the field of computer science, software
development or a related field". See the official rules for details of eligibility rules and the timetable as well as the contest description. As well as cash prizes, Microsoft will contest participants who have entered the live evaluation competitions opportunity to have their entries evaluated for the potential award of an intellectual property agreement with Microsoft, for which Microsoft may offer two agreements of up to $100,000 each.

Having recently announced its Kinect Accelerator program Microsoft really does seem keen on making the Kinect a serious commercial success.