Matjaž Jogan: Visual categorization

The state of the art methods for visual categorization typically use supervised or unsupervised visual learning to construct categorical representations which discriminate between object categories and background. In our framework, we investigate an alternative approach, which uses hierarchical synchronous feature binding and matching to find high level structural and appearance similarities. We will present ongoing work and experiments on category matching in 2D images and discuss a possible extension to categorization in 3D.

Matjaž Jogan is a member of the Visual Cognitive Systems Laboratory and  the Computer Vision Laboratory at the University of Ljubljana. His main research field is hierarchical visual learning and categorization for object and place categorization.