2017 Idiap Awards for James Newling and Gulcan Can

Idiap PhD Student Research Award

James' research is in the field of clustering (k-means and k-medoids). James Newling started in September 2013 and defends his thesis in December 2017. His PhD advisor, François Fleuret, says: "While his work is strongly math-oriented, with proven complexity bounds, [James] also did a fantastic work regarding implementation. He spent three months as an intern at AMD in Texas last year to develop new algorithms for general matrix products on their line of GPU, and it went so well that AMD [extended support] so that we could continue working on that topic." James has several publications, at NIPS (x2), ICML and AISTATS. The AISTATS paper got the best paper award.

Idiap PhD Student Paper Award

Gulcan Can, J.-M. Odobez, and D. Gatica-Perez "Codical Maya Glyph Segmentation: A Crowdsourcing Approach," IEEE Trans. on Multimedia, published online Sep. 2017; doi: 10.1109/TMM.2017.2755985

This paper, published online in September 2017 in the IEEE transaction on Multimedia, focuses on the crowd-annotation of an ancient Maya glyph dataset derived from the three ancient codices that survived up to date. It leverages non-expert annotators to segment glyph-blocks into their constituent glyph. Through design of an engaging task, taking into account crowd behavior and providing supervision during the task, a unique dataset of over 9000 Maya glyphs from 291 categories individually segmented from the three codices was created and will be made publicly available thanks to this process. Gulcan has publicly defended her thesis in November 2017.