CoChat
Sign Up
Cited by 1,269Visual perception and processing mechanismsBayesian Modeling and Causal InferenceFace Recognition and Perception

Object perception as Bayesian inference

Open access
Kersten, Daniel, Mamassian, Pascal, Yuille, Alan·2004·Annual Review of Psychology
DOI 10.1146/annurev.psych.55.090902.142005
Open this paper in CoChat — free accountOpen PDF

Abstract

We perceive the shapes and material properties of objects quickly and reliably despite the complexity and objective ambiguities of natural images. Typical images are highly complex because they consist of many objects embedded in background clutter. Moreover, the image features of an object are extremely variable and ambiguous owing to the effects of projection, occlusion, background clutter, and illumination. The very success of everyday vision implies neural mechanisms, yet to be understood, that discount irrelevant information and organize ambiguous or noisy local image features into objects and surfaces. Recent work in Bayesian theories of visual perception has shown how complexity may be managed and ambiguity resolved through the task-dependent, probabilistic integration of prior object knowledge with image features.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1146/annurev.psych.55.090902.142005

Cites

158 of 158 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The GNW theory can be combined with Bayesian inference, in which GNW neurons provide a compressed high-level model of the external world, top-down signals broadcast predictions, and bottom-up signals convey sensory input and either amplify predicted signals or measure prediction error, allowing the central model to be updated and achieving a coherent multisensory model.

    “GNW neurons, lying at the top of a deep feedforward network, provide a compressed, high-level symbolic model of an aspect of the external world; top-down signals broadcast the predictions that this model makes to lower-level areas; and bottom-up signals convey sensory signals and may either amplify the signals predicted by top-down input”

    From Conscious Processing and the Global Neuronal Workspace Hypothesis (2020) · cites this paper as Kersten et al., 2004

    Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

    Experimental — accuracy not yet measured

  • The posterior distribution, formed by multiplying likelihood and prior, strongly favors some scene interpretations over others.

    “The product of the likelihood and prior-the posterior distribution over scenes, given the image y 0 -thus strongly favors some scene interpretations over others”

    From The Interface Theory of Perception (2015) · cites this paper as Kersten, Mamassian & Yuille, 2004

  • Human observers often approach an ideal observer, inspiring the idea that the human perceptual system may implement approximate Bayesian inference.

    “Interestingly, human observers often approach such an 'ideal observer' ([102], but see [103]), inspiring the idea that the human perceptual system may implement (approximate) Bayesian inference [104].”

    From How Do Expectations Shape Perception? (2018) · cites this paper as [104]

    Experimental — accuracy not yet measured

  • The two mechanisms referred to as 'shutting up' and 'stop gossiping' are both accommodated within predictive coding and operate on distinct populations (error and prediction units, respectively).

    “These two mechanisms, which have been referred to as 'shutting up' and 'stop gossiping' [4,104] (see also previous paragraph on dampening vs sharpening) are both accommodated within predictive coding, and operate on distinct populations (error and prediction units, respectively).”

    From How Do Expectations Shape Perception? (2018) · cites this paper as [104]

  • Disentangling the many external states that could underlie any given sensory input is an ill-posed inverse problem that requires constraints or regularization.

    “Disentangling the many external states that could underlie any given sensory input is an ill-posed inverse problem that requires constraints or regularization (e.g., by the priors of a generative model; Lee and Mumford, 2003; Kersten et al., 2004).”

    From Allostatic Self-efficacy: A Metacognitive Theory of Dyshomeostasis-Induced Fatigue and Depression (2016) · cites this paper as Kersten et al., 2004

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

1 of the 159 references couldn’t be matched to a known paper.

Research works printed but not linked: 4

  1. [52]52. Kanizsa G, & Gerbino W. 1976. Convexity and symmetry in figure-ground organisation. In Vision and Artifact, ed. M Henle. New York: Springer
  2. [61]61. Knill DC, & Richards W. 1996. Perception as Bayesian Inference. Cambridge: Cambridge Uni- versity Press
  3. [124]124. Tu Z, & Zhu S-C. 2002. Parsing Images into Region and Curve Processes. Presented at Proc. of the 7th European Conference on Computer Vision
  4. [125]125. Tu Z, Chen A, Yuille AL, & Zhu SC. 2003. Image Parsing. Proceedings of International Confer- ence on Computer Vision. ICCV'2003. Cannes. France.

Cited by

Sorted by citations

Related

Authors
Kersten, D., Mamassian, P., & Yuille, A.
Affiliations
  • University of Minnesota (US)
  • University of Glasgow (GB)
  • University of California, Los Angeles (US)
ORCID
Venue
Annual Review of Psychology
Publisher
Annual Reviews
Published
2004-02-01
Volume
55
Issue
1
Pages
271-304
ISSN
  • 0066-4308 (ISSN-L)
  • 1545-2085
DOI
10.1146/annurev.psych.55.090902.142005
OpenAlex
W1497935377
Cited by
1,269
References
159
Citations by year
2026: 33 · 2025: 55 · 2024: 45 · 2023: 49 · 2022: 63 · 2021: 53 · 2020: 71 · 2019: 97 · 2018: 71 · 2017: 60 · 2016: 78 · 2015: 70 · 2014: 69 · 2013: 57 · 2012: 52
FWCI
3.33
Type
journal article
Language
en
Open access
Locations
  • Annual Review of Psychology (publishedVersion)
  • PubMed (publishedVersion)
  • eScholarship (California Digital Library) (submittedVersion)
  • Department of Statistics, UCLA
Funders
  • National Institutes of Health (5r01ey011507-04, 5r01ey012691-03)
  • National Eye Institute (R01EY013875, EY12691, R01 EY11507-001, EY013875, EY02587)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Visual perception and processing mechanisms
  • Physical Sciences › Computer Science › Artificial Intelligence › Bayesian Modeling and Causal Inference
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Face Recognition and Perception
Keywords
object perception, Bayesian inference, prior knowledge, probabilistic integration, object shape perception, occlusion, background clutter, illumination effects
MeSH terms
Bayes Theorem, Decision Making, Humans, Learning, Psychological Theory, Psychophysics, Visual Perception

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review