Context-based vision system for place and object recognition

  • Authors:
  • Antonio Torralba;Kevin P. Murphy;William T. Freeman;Mark A. Rubin

  • Affiliations:
  • -;-;-;-

  • Venue:
  • ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
  • Year:
  • 2003

Quantified Score

Hi-index 0.02

Visualization

Abstract

While navigating in an environment, a vision system has to be ableto recognize where it is and what the main objects in the sceneare. In this paper we present a context-based vision system forplace and object recognition. The goal is to identify familiarlocations (e.g., office 610, conference room 941, Main Street), tocategorize new environments (office, corridor, street) and to usethat information to provide contextual priors for objectrecognition (e.g., tables are more likely in an office than astreet). We present a low-dimensional global image representationthat provides relevant information for place recognition andcategorization, and show how such contextual information introducesstrong priors that simplify object recognition. We have trained thesystem to recognize over 60 locations (indoors and outdoors) and tosuggest the presence and locations of more than 20 different objecttypes. The algorithm has been integrated into a mobile system thatprovides real-time feedback to the user.