More about me - Research

Click here if you were searching for my contact data.

Adrian ION

I am a postdoctoral researcher affiliated with PRIP and IST Austria.

This (click to download) is a more recent CV.

My research experience includes:

• low and mid-level image segmentation
• (structured-) model based tracking
• object recognition
• eccentricity (shape) transform and shape matching (topic of my PhD thesis)
• (hierarchical) computation of topological invariants
• hierarchical combinatorial representations

My current research focuses on structured models (learning and inference) for computer vision tasks.

Recent results from this line of work include the state-of-the-art image segmentation method FGTiling [ICS, ICCV 2011], and a statistical model for joint segmentation and labeling JSL [ICS, NIPS 2011], which was a winning entry (named "BONN_FGT_SEGM") of the Pascal VOC 2011 Semantic Segmentation Challenge. Below are two images labeled by JSL using FGTiling to sample segmentations.

My master thesis was on photorealistic rendering (Monte Carlo Path Tracing).
During my high-school and student times I had a passion for programming problems.