Cancelled.





Carsten Rother: CRFs in Action: Decision Tree Fields and Intrinsic Images
In this talk I will present two different pieces of work, which were presented
recently at NIPS '11, ICCV '11 and CVPR '12, which both utilize the power of
Conditional Random Fields (CRFs).

Decision Tree Fields (DTF) and Regression Tree Fields (RTFs) are new models
that
combine and generalize random forests and conditional random fields (CRF). The
key idea is to have a very large number of potential functions which are based
on non-parametric decision trees or regression trees. Although the resulting
model has millions of parameters, it is tractable to perform efficient learning
and inference with it. We demonstrate excellent performance for various tasks
such as in-painting, de-noising, and person detection in depth images.

The task of recovering intrinsic images is to separate a given input image into
its material-dependent properties, known as reflectance or albedo, and its
light-dependent properties, such as shading and shadows. We develop a new CRF
model which achieves state-of the results. The key novel ingredient is a
sparseness prior on reflectance, which encodes the property that a scene is
often composed of a few different materials.