This paper introduces an approach for dense 3D reconstruction from unregistered Internet-scale photo collections with about 3 million of images within the span of a day on a single...
The paper presents a novel coding technique based on approximate geometry for images taken from arbitrary recording positions around a 3-D scene. Such data structures occur in ima...
Structured outputs such as multidimensional vectors or graphs are frequently encountered in real world pattern recognition applications such as computer vision, natural language pr...
This paper proposes a method to recover the embedding
of the possible shapes assumed by a deforming nonrigid
object by comparing triplets of frames from an orthographic
video se...
Vincent Rabaud (University of California, San Dieg...
High-level, or holistic, scene understanding involves
reasoning about objects, regions, and the 3D relationships
between them. This requires a representation above the
level of ...