Abstract. Most current approaches to recognition aim to be scaleinvariant. However, the cues available for recognizing a 300 pixel tall object are qualitatively different from tho...
We consider the problem of segmenting an image into foreground and background, with foreground containing solely objects of interest known a priori. We propose an integration mode...
We present a simple framework to model contextual
relationships between visual concepts. The new framework
combines ideas from previous object-centric methods
(which model conte...
Nikhil Rasiwasia (University Of California, San Di...
Given very few images containing a common object of interest under severe variations in appearance, we detect the common object and provide a compact visual representation of that...
Shai Bagon, Or Brostovski, Meirav Galun, Michal Ir...
Inspection is considered a powerful method to check software documents for defects. Many published work shows that inspections in requirements specification phase are particularly...