The winning strategy proposed by LIRIS is a result of joint work of

• Natalia Neverova (LIRIS),
• Christian Wolf (LIRIS), 
• Graham Taylor (University of Guelph, Canada),
• Florian Nebout (Awabot).

The objective of the challenge was to detect, localize and classify Italian conversational gestures from large database of 13858 gestures. The multimodal data included color video, range maps and a skeleton stream.

This year, the challenge attracted near 200 participants from different institutions. 17 teams have submitted their codes and predictions on the final evaluation dataset of the given track.

The winning entry is based on a deep learning architecture with spatial gesture decomposition into large-scale body motion and fine finger articulation. The idea of learning at multiple scales is also applied to the temporal dimension. Fusing multiple modalities at several spatial and temporal scales lead to a significant increase in recognition rates, allowing the model to compensate for errors of the individual classifiers as well as noise in the separate channels. 

The exact strategy will be presented in September 2014 at the ECCV workshop dedicated to the challenge. A journal paper is under writing.

Results: http://gesture.chalearn.org/challenge-results

• This work was partially funded by french grant INTERABOT, a call "Investissements d'Avenir / Briques Génériques du Logiciel Embarqué"
• This work was partially funded by the french region "Rhônes Alpes" through the ARC6 research council.