Models of object recognition

doi:10.1016/0959-4388(91)90089-P

Current Opinion in Neurobiology

Volume 1, Issue 2, August 1991, Pages 270-273

https://doi.org/10.1016/0959-4388(91)90089-P Get rights and content

Abstract

Progress in the understanding of visual recognition in the past year has been signified by the demonstration of computational feasibility of and psychophysical support for two-dimensional view-interpolation methods.

References (41)

E. Rosch et al.
Basic Objects in Natural Categories
Cogn Psychol
(1976)
S. Ullman
Aligning Pictorial Descriptions: an Approach to Object Recognition
Cognition
(1989)
D.G. LowE
Three-Dimensional Object Recognition from Single Two-Dimensional Images
Artif Intell
(1987)
S. Edelman et al.
Viewpoint-Specific Representa ..tions in 3D Object Recognition
M. Mishkin et al.
Object Vision and Spatial Vision: Two Cortical Pathways
Trends Neurosci
(1983)
H.B. Barlow
The Role of Single Neurons in the Psychology of Perception
Q J Exp Psychol (A)
(1985)
T. Nazir et al.
Some Results on Translation Invariance in the Human Visual System
Spatial Vis
(1990)
D. Mark
Vision [book]
(1982)
T. Poggio et al.
Parallel Integration of Vision Modules
Science
(1988)
D.W. Thompson et al.
Three-Dimensional Model Matching from an Unconstrained Viewpoint

D.G. Lowe

Stabilized Solution for 3D Model Parameters

S. Edelman et al.

Computational Vision: a Critical Review

S. Ullman et al.

Recognition by Linear Combinations of Models

T. Poggio et al.

A Network that Learns to Recognize Three-Dimensional Objects

Nature

(1990)

D.S. Broomhead et al.

Multivariable Functional Interpolation and Adaptive Networks

Complex Syst

(1988)

T. Poggio et al.

Regularization Algorithms for Learning that Are Equivalent to Multilayer Networks

Science

(1990)

S. Edelman et al.

Bringing the Grandmother Back into the Picture: a Memory-Based View of Object Recognition

S.E. Palmer et al.

Canonical Perspective and the Perception of Objects

M. Tarr et al.

Mental Rotation and Orientation-Dependence in Shape Recognition

Cogn Psychol

(1989)

S. Edelman et al.

Stimulus Familiarity Determines Recognition Strategy for Novel 3D Objects

Cited by (16)

Recognition of rotated objects and cognitive offloading in dogs
2022, iScience
Citation Excerpt :
Several alternative models have been proposed to explain how the human visual system recognizes whether two objects seen from different points of view are the same or not. For example, some researchers proposed that object recognition might be based on more than just one processing mechanism (the “multiple routes” hypothesis, Vanrie et al. (2001)) or that it might be based on interpolation between the limited views of an object stored in memory (Edelman and Poggio, 1991; Riesenhuber and Poggio, 2000). The strategies used by non-human species to recognize rotated objects are debated too, as discussed below.
Recognition of rotated images can challenge visual systems. Humans often diminish the load of cognitive tasks employing bodily actions (cognitive offloading). To investigate these phenomena from a comparative perspective, we trained eight dogs (Canis familiaris) to discriminate between bidimensional shapes. We then tested the dogs with rotated versions of the same shapes, while measuring their accuracy and head tilts. Although generalization to rotated stimuli challenged dogs (overall accuracy: 55%), three dogs performed differently from chance level with rotated stimuli. The amplitude of stimulus rotation did not influence dogs’ performance. Interestingly, dogs tilted their head following the direction and amplitude of rotated stimuli. These small head movements did not influence their performance. Hence, we show that dogs might be capable of recognizing rotated 2D objects, but they do not use a cognitive offloading strategy in this task. This work paves the way to further investigation of cognitive offloading in non-human species.
Mechanosensation and Adaptive Motor Control in Insects
2016, Current Biology
Citation Excerpt :
One way to identify compelling scientific problems is to ask which features of biological systems are most difficult to replicate in artificial systems. Thirty years ago, a reasonable answer might have been the problem of visual object recognition, which motivated many neurophysiological and computational studies of the primate visual cortex [195]. However, recent advances in artificial neural networks have made it possible for a computer to automatically classify natural images with accuracies that match, and sometimes surpass, human performance [196].
The ability of animals to flexibly navigate through complex environments depends on the integration of sensory information with motor commands. The sensory modality most tightly linked to motor control is mechanosensation. Adaptive motor control depends critically on an animal’s ability to respond to mechanical forces generated both within and outside the body. The compact neural circuits of insects provide appealing systems to investigate how mechanical cues guide locomotion in rugged environments. Here, we review our current understanding of mechanosensation in insects and its role in adaptive motor control. We first examine the detection and encoding of mechanical forces by primary mechanoreceptor neurons. We then discuss how central circuits integrate and transform mechanosensory information to guide locomotion. Because most studies in this field have been performed in locusts, cockroaches, crickets, and stick insects, the examples we cite here are drawn mainly from these ‘big insects’. However, we also pay particular attention to the tiny fruit fly, Drosophila, where new tools are creating new opportunities, particularly for understanding central circuits. Our aim is to show how studies of big insects have yielded fundamental insights relevant to mechanosensation in all animals, and also to point out how the Drosophila toolkit can contribute to future progress in understanding mechanosensory processing.
View-invariant object category learning, recognition, and search: How spatial and object attention are coordinated using surface-based attentional shrouds
2009, Cognitive Psychology
Citation Excerpt :
One problem with such theories is that one can easily find objects that do not seem to conform to any pre-determined set of parts. According to “view-based” object recognition theories, objects are represented as collections of view-specific representations, leading to recognition performance that is a function of previously seen object views (Bulthoff & Edelman, 1992; Edelman & Poggio, 1991; Tarr & Bulthoff, 1995, 1998; Tarr, Williams, Hayward, & Gauthier, 1998). As noted above, ARTSCAN employs a view-based object learning and recognition strategy.
How does the brain learn to recognize an object from multiple viewpoints while scanning a scene with eye movements? How does the brain avoid the problem of erroneously classifying parts of different objects together? How are attention and eye movements intelligently coordinated to facilitate object learning? A neural model provides a unified mechanistic explanation of how spatial and object attention work together to search a scene and learn what is in it. The ARTSCAN model predicts how an object’s surface representation generates a form-fitting distribution of spatial attention, or “attentional shroud”. All surface representations dynamically compete for spatial attention to form a shroud. The winning shroud persists during active scanning of the object. The shroud maintains sustained activity of an emerging view-invariant category representation while multiple view-specific category representations are learned and are linked through associative learning to the view-invariant object category. The shroud also helps to restrict scanning eye movements to salient features on the attended object. Object attention plays a role in controlling and stabilizing the learning of view-specific object categories. Spatial attention hereby coordinates the deployment of object attention during object category learning. Shroud collapse releases a reset signal that inhibits the active view-invariant category in the What cortical processing stream. Then a new shroud, corresponding to a different object, forms in the Where cortical processing stream, and search using attention shifts and eye movements continues to learn new objects throughout a scene. The model mechanistically clarifies basic properties of attention shifts (engage, move, disengage) and inhibition of return. It simulates human reaction time data about object-based spatial attention shifts, and learns with 98.1% accuracy and a compression of 430 on a letter database whose letters vary in size, position, and orientation. The model provides a powerful framework for unifying many data about spatial and object attention, and their interactions during perception, cognition, and action.
An optimal estimation approach to visual perception and learning
1999, Vision Research
How does the visual system learn an internal model of the external environment? How is this internal model used during visual perception? How are occlusions and background clutter so effortlessly discounted for when recognizing a familiar object? How is a particular object of interest attended to and recognized in the presence of other objects in the field of view? In this paper, we attempt to address these questions from the perspective of Bayesian optimal estimation theory. Using the concept of generative models and the statistical theory of Kalman filtering, we show how static and dynamic events occurring in the visual environment may be learned and recognized given only the input images. We also describe an extension of the Kalman filter model that can handle multiple objects in the field of view. The resulting robust Kalman filter model demonstrates how certain forms of attention can be viewed as an emergent property of the interaction between top–down expectations and bottom–up signals. Experimental results are provided to help demonstrate the ability of such a model to perform robust segmentation and recognition of objects and image sequences in the presence of occlusions and clutter.
Computational approaches to cognition: the bottom-up view
1993, Current Opinion in Neurobiology
Active vision: On the relevance of a bio-inspired approach for object detection
2020, Bioinspiration and Biomimetics

View all citing articles on Scopus

View full text

Models of object recognition

Abstract

Cogn Psychol

Cognition

Artif Intell

Trends Neurosci

Q J Exp Psychol (A)

Spatial Vis

Vision [book]

Parallel Integration of Vision Modules

Science

Three-Dimensional Model Matching from an Unconstrained Viewpoint

Stabilized Solution for 3D Model Parameters

Computational Vision: a Critical Review

Recognition by Linear Combinations of Models

A Network that Learns to Recognize Three-Dimensional Objects

Nature

Multivariable Functional Interpolation and Adaptive Networks

Complex Syst

Regularization Algorithms for Learning that Are Equivalent to Multilayer Networks

Science

Bringing the Grandmother Back into the Picture: a Memory-Based View of Object Recognition

Canonical Perspective and the Perception of Objects

Mental Rotation and Orientation-Dependence in Shape Recognition

Cogn Psychol

Stimulus Familiarity Determines Recognition Strategy for Novel 3D Objects