In this talk we present our approach to visual categorization which combines a learned compositional hierarchy, representing shapes of multiple object classes, and a coarse-to-fine matching scheme that exploits a taxonomy of objects to perform efficient object detection. From simple contour fragments frequent spatial configurations are learnt and recursively combined into increasingly more complex and class-specific shape compositions layer by layer. To speed up recognition, a taxonomy of constellation models cascaded from coarse-to-fine resolution is constructed automatically in a way that minimizes the number of expected computations during recognition by optimizing the cost-to-power ratio.
B.04 Guest lecture
COBISS.SI-ID: 8991316We investigated the methods that can be used in automatic video surveillance systems and the most appropriate ones were used in the implementation. The prototype was developed on BeagleBoard with Angstrom Linux platform due to its similarity with mobile target platforms. The implementation was mainly based on the Motion software package supported by the driver library Video for Linux (V4L).
F.17 Transfer of existing technologies, know-how, methods and procedures into practice
COBISS.SI-ID: 8907348Miha Moškon was invited to Technical University of Varna as a speaker. The lecture was given in Varna, Bulgaria, on September 23, 2010.
B.04 Guest lecture
COBISS.SI-ID: 8179284In the University of Ljubljana's 'Week of University' we presented a demonstration 'Intelligent bus stop'. The presentation took place in Faculty of computer and information science, in Ljubljana, on December 1, 2010.
F.28 Organising an exhibition
COBISS.SI-ID: 8098388Aleš Leonardis was invited to the "SIG-09: First International Workshop on Stochastic Image Grammars" as a keynote speaker. The workshop was held in Miami, Florida, on June 21, 2009. http://civs.stat.ucla.edu/sig09/keynotespeakers.html
B.04 Guest lecture
COBISS.SI-ID: 7460692