Current approaches to object category recognition require datasets of training images to be manually prepared, with varying degrees of supervision. We present an approach that can...
Robert Fergus, Fei-Fei Li 0002, Pietro Perona, And...
In real-world machine learning problems, it is very common that part of the input feature vector is incomplete: either not available, missing, or corrupted. In this paper, we pres...
: Contour Estimation, Bayesian Estimation, Random Fields, Dynamic Programming, Multigrid Methods. This paper addresses contour estimation on images modeled as piecewise homogeneous...
Several operations of Web-based applications are optimized with respect to the set of resources that will receive the majority of requests in the near future, namely the hot set. U...
Software system documentation is almost always expressed informally in natural language and free text. Examples include requirement specifications, design documents, manual pages, ...