Map Building from Human-Computer Interactions
Abstract
On-line help from a human actor will be exploited to facilitate computer perception. This paper proposes and innovative real-time algorithm - running on an active vision head - to build 3D scene descriptions from human cues. The theory is supported by experimental results both for figure/ground segregation of typical heavy objects in a scene (such as furniture), and for 3D object/scene reconstruction.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2004
- Accession Number
- ADA434764
Entities
Organizations
- Massachusetts Institute of Technology