Object perception: Generative image models and bayesian inference

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Scopus citations


Humans perceive object properties such as shape and material quickly and reliably despite the complexity and objective ambiguities of natural images. The visual system does this by integrating prior object knowledge with critical image features appropriate for each of a discrete number of tasks. Bayesian decision theory provides a prescription for the optimal utilization of knowledge for a task that can guide the possibly sub-optimal models of human vision. However, formulating optimal theories for realistic vision problems is a non-trivial problem, and we can gain insight into visual inference by first characterizing the causal structure of image features-the generative model. I describe some experimental results that apply generative models and Bayesian decision theory to investigate human object perception.

Original languageEnglish (US)
Title of host publicationBiologically Motivated Computer Vision - 2nd International Workshop, BMCV 2002, Proceedings
EditorsHeinrich H. Bulthoff, Christian Wallraven, Seong-Whan Lee, Tomaso A. Poggio
PublisherSpringer Verlag
Number of pages12
ISBN (Electronic)9783540001744
StatePublished - 2002
Event2nd International Workshop on Biologically Motivated Computer Vision, BMCV 2002 - Tubingen, Germany
Duration: Nov 22 2002Nov 24 2002

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Other2nd International Workshop on Biologically Motivated Computer Vision, BMCV 2002


Dive into the research topics of 'Object perception: Generative image models and bayesian inference'. Together they form a unique fingerprint.

Cite this