The problem of multibody motion segmentation is an important and challenging issue in computer vision. In this paper, a novel segmentation technique based on simulated annealing (...
In a known environment, objects may be tracked in multiple views using a set of background models. Stereo-based models can be illumination-invariant, but often have undefined valu...
Trevor Darrell, David Demirdjian, Neal Checka, Ped...
Sliding window classifiers are among the most successful and widely applied techniques for object localization. However, training is typically done in a way that is not specific to...
Recent work both in the relational and the XML world have shown that the efficacy and efficiency of duplicate detection is enhanced by regarding relationships between entities. Ho...
When constructing a classifier from labeled data, it is important not to assign too much weight to any single input feature, in order to increase the robustness of the classifier....