In this paper we reformulate the 3D reconstruction of deformable surfaces from monocular video sequences as a labeling problem. We solve simultaneously for the assignment of featu...
Abstract. Models for concurrency can be classified with respect to three relevant parameters: behaviour/system, interleaving/noninterleaving, linear/branching time. When modelling...
To track objects in video sequences, many studies have been done to characterize the target with respect to its color distribution. Most often, the Gaussian Mixture Model (GMM) is ...
We describe a directed bilinear model that learns higherorder groupings among features of natural images. The model represents images in terms of two sets of latent variables: one...
Jack Culpepper, Jascha Sohl-Dickstein, Bruno Olaha...
A novel video object segmentation method is proposed which aims at combining color and motion information. The model has a multilayer structure: Each feature has its own layer, cal...