Brief Bio

Adrian Ion has received an engineer's degree in computer science, from the “Politehnica” University of Timisoara, Romania, in 2001, and a doctorate in Computer Science with a focus on computer vision, from the Vienna University of Technology, in 2009 (supervisor Prof. Walter G. Kropatsch). He has done postdoctoral research in the Computer Vision and Machine Learning group of the University of Bonn (supervisor Prof. Cristian Sminchisescu), the Vienna University of Technology, and the Institute of Science and Technology Austria. Work by himself and his collaborators was a winner of the PASCAL VOC Object Segmentation Challenge in 2011 and 2012. He is a recipient of the Best Paper Award of the Austrian Association for Pattern Recognition, in 2006. Prior to that, he received a number of prizes at national (Romanian) and international programming contests. He has done reviews for major conferences and journals in the field (ICCV, ECCV, NIPS, PAMI). He is currently with the Imagination Computer Services GmbH in Vienna.

Interests

My research experience includes image segmentation and labeling, object tracking, shape matching, as well as aspects of structured learning and inference, graph algorithms, and computational topology.

My interests are in structured models (learning and inference) for computer vision tasks.

Results from this line of work include the FGTiling mid-level image segmentation method [PDF], and building on it, the JSL statistical model for joint segmentation and labeling [PDF], which scored 47.5% on the Pascal VOC Object Segmentation Challenge 2012 (additional components used: Figure-Ground segment proposals: CPMC, predictors for sequential Figure-Ground ranking, and the O2P feature).

Below are example output semantic-segmentations produced by JSL (the inferred regions are outlined in white).