Two members of the project group (Matej Kristan and Luka Čehovin) organized the 17th computer vision winter workshop (CVWW2012). The workshop connects researchers from Austria, Slovenia and Chechz Republic, that work in the field of computer vision. At this workshop we have presented preliminary results from the project ARRS-RPROJ-LP2010/ 103 to a wider audience of researchers from computer vision.
B.01 Organiser of a scientific meeting
COBISS.SI-ID: 259992320Three members of the project group (Matej Kristan, Luka Čehovin and Aleš Leonardis) coorganized the first challenge in short-term visual tracking VOT2013. The organization committee was international, spreading groups from Slovenia, UK, Austria, Czech Republic and Australia. The challenge was accompanied by a workshop organized at the top computer vision conference ICCV2013. The workshop was well as the challenge were well accepted by the community. Within the challenge, we have compared 27 diverse approaches to tracking of arbitrary, articulated and nonarticulated objects, using a new annotated dataset and new performance evaluation methodology. A paper with the challenge results, coauthored by ~50 authors, was published and presented at the workshop.
B.01 Organiser of a scientific meeting
COBISS.SI-ID: 10409300Visual categorisation has been an area of intensive research in the vision community for several decades. Ultimately, the goal is to efficiently detect and recognize an increasing number of object classes. The problem entangles three highly interconnected issues: the internal object representation, which should compactly capture the visual variability of objects and generalize well over each class; a means for learning the representation from a set of input images with as little supervision as possible; and an effective inference algorithm that robustly matches the object representation against the image and scales favorably with the number of objects. In this talk I presented our novel approach which combines a learned compositional hierarchy, representing (2D) shapes of multiple object classes, and a coarse-to-fine matching scheme that exploits a taxonomy of objects to perform efficient object detection. The combination of the learned taxonomy with the compositional hierarchy of object shape achieves efficiency both with respect to the representation of the structure of objects and in terms of the number of modeled object classes. The experimental results show that the learned multi-class object representation achieves a detection performance comparable to the current state-of-the-art flat approaches with both faster inference and shorter training times.
B.04 Guest lecture
COBISS.SI-ID: 8991316Visual categorisation has been an area of intensive research in the vision community for several decades. Ultimately, the goal is to efficiently detect and recognize an increasing number of object classes. The problem entangles three highly interconnected issues: the internal object representation, which should compactly capture the visual variability of objects and generalize well over each class; a means for learning the representation from a set of input images with as little supervision as possible; and an effective inference algorithm that robustly matches the object representation against the image and scales favorably with the number of objects. In this talk I presented our novel approach which combines a learned compositional hierarchy, representing (2D) shapes of multiple object classes, and a coarse-to-fine matching scheme that exploits a taxonomy of objects to perform efficient object detection. The combination of the learned taxonomy with the compositional hierarchy of object shape achieves efficiency both with respect to the representation of the structure of objects and in terms of the number of modeled object classes. The experimental results show that the learned multi-class object representation achieves a detection performance comparable to the current state-of-the-art flat approaches with both faster inference and shorter training times.
B.04 Guest lecture
COBISS.SI-ID: 10487124