Representations for Cognitive Processes

01.01.2007 - 15.12.2009
Forschungsförderungsprojekt
Project Title: Cognitive Vision - Key Technology for Personal Assistance (http://fsp.acin.tuwien.ac.at/) Subproject 3: Representations for Cognitive Vision (SP3) The subproject "Representations for Cognitive Vision" provides new and improves existing representations that are useful for cognitive vision. It became obvious in the first part of the project (2004-2006) that not a single representation is suitable for all tasks. For generic vision tasks there is the necessity for complementary representations. In particular, both generative (reconstructive) and discriminative representations are needed. This is required for more autonomous learning and for obtaining robustness in general. In a similar spirit a cognitive vision system needs both global (holistic) and local (part-based) representations. In the second half of the project, SP3 plans to build on the results achieved within the first period of the research project. We continue along our three major lines of research in appearance-based, spatio-temporal, and graph-based representations. As a major application in SP3, we will test our newly developed representations on specific and generic object recognition tasks. PRIP (http://www.prip.tuwien.ac.at/Research/FSPCogVis/), Task SP3.3 - Graph-based representations: We will address representation and abstraction in cognitive vision from the local, object level, to the global, scene level, all with a special care for the dynamic aspects, concentrating on (i) dynamic skeletons to represent moving shapes, (ii) invariants using homology groups, (iii) topology preserving description of dynamic scenes, and (iv) landmark-based addressing schemes to link topology with metric properties.

Personen

Projektleiter_in

Projektmitarbeiter_innen

Institut

Grant funds

  • FWF - Österr. Wissenschaftsfonds (National) Austrian Science Fund (FWF)

Forschungsschwerpunkte

  • Media Informatics and Visual Computing: 70%
  • Mathematical and Algorithmic Foundations: 20%
  • Computer Science Foundations: 10%

Schlagwörter

DeutschEnglisch
Kognitive maschinelles Sehencognitive computer vision
lokale + globale Repräsentationenlocal + global representations
Visuelle Abstraktionvisual abstraction
Graph-basierte Repräsentationengraph based representations

Publikationen