Face and Object Recognition Using Biological Features and Few Views

In recent years, a large number of impressive face and object recognition algorithms have surfaced, both computational and biologically inspired. Only a few of these can detect face and object views. Empirical studies concerning face and object recognition suggest that faces and objects may be stored in our memory by a few canonical representations. In cortical area V1 exist double-opponent colour blobs, also simple, complex, and end-stopped cells that provide input for a multiscale line and edge representation, keypoints for dynamic feature routing, and saliency maps for Focus-of-Attention. All these combined allow us to segregate faces. Events of different facial views are stored in memory and combined in order to identify the view and recognise a face, including its expression. The authors show that with five 2D views and their cortical representations it is possible to determine the left-right and frontal-lateral-profile views, achieving view-invariant recognition. They also show that the same principle with eight views can be applied to 3D object recognition when they are mainly rotated about the vertical axis. Although object recognition is here explored as a special case of face recognition, it should be stressed that faces and general objects are processed in different ways in the cortex.

Download Full-text

Cortical 3D Face and Object Recognition Using 2D Projections

International Journal of Creative Interfaces and Computer Graphics ◽

10.4018/jcicg.2012010104 ◽

2012 ◽

Vol 3 (1) ◽

pp. 45-62 ◽

Cited By ~ 3

Author(s):

João Rodrigues ◽

Roberto Lam ◽

Hans du Buf

Keyword(s):

Object Recognition ◽

Cortical Area ◽

Empirical Studies ◽

Recognition Rate ◽

Vertical Axis ◽

Dynamic Feature ◽

Simple Complex ◽

Saliency Maps ◽

Canonical Representations ◽

Cortical Representations

Empirical studies concerning face recognition suggest that faces may be stored in memory by a few canonical representations. In cortical area V1 exist double-opponent colour blobs, also simple, complex and end-stopped cells which provide input for a multiscale line/edge representation, keypoints for dynamic feature routing, and saliency maps for Focus-of-Attention. All these combined allow faces to be segregated. Events of different facial views are stored in memory and combined to identify the view and recognise a face, including its expression. In this paper, the authors show that with five 2D views and their cortical representations it is possible to determine the left-right and frontal-lateral-profile views, achieving a view-invariant recognition rate of 91%. The authors also show that the same principle with eight views can be applied to 3D object recognition when they are mainly rotated about the vertical axis.

Download Full-text

Visual Behavior Based Bio-Inspired Polarization Techniques in Computer Vision and Robotics

Developing and Applying Biologically-Inspired Vision Systems ◽

10.4018/978-1-4666-2539-6.ch011 ◽

2012 ◽

pp. 243-272 ◽

Cited By ~ 2

Author(s):

Abd El Rahman Shabayek ◽

Olivier Morel ◽

David Fofi

Keyword(s):

Computer Vision ◽

Object Recognition ◽

Visual Perception ◽

Polarization Vision ◽

Visual Behavior ◽

Biologically Inspired ◽

Long Time ◽

Comprehensive Survey ◽

High Level ◽

And Robotics

For long time, it was thought that the sensing of polarization by animals is invariably related to their behavior, such as navigation and orientation. Recently, it was found that polarization can be part of a high-level visual perception, permitting a wide area of vision applications. Polarization vision can be used for most tasks of color vision including object recognition, contrast enhancement, camouflage breaking, and signal detection and discrimination. The polarization based visual behavior found in the animal kingdom is briefly covered. Then, the authors go in depth with the bio-inspired applications based on polarization in computer vision and robotics. The aim is to have a comprehensive survey highlighting the key principles of polarization based techniques and how they are biologically inspired.

Download Full-text

Basic Geometric Dispersion Theory of Decision Making Under Risk: Asymmetric Risk Relativity, New Predictions of Empirical Behaviors, and Risk Triad

Decision Analysis ◽

10.1287/deca.2019.0404 ◽

2021 ◽

Author(s):

Behnam Malakooti ◽

Mohamed Komaki ◽

Camelia Al-Najjar

Keyword(s):

Risk Behaviors ◽

Empirical Studies ◽

Dispersion Theory ◽

Dispersion Models ◽

Decision Making Under Risk ◽

First Order ◽

Order Stochastic Dominance ◽

Out Of Sample ◽

Mean Variance ◽

Special Case

Many studies have spotlighted significant applications of expected utility theory (EUT), cumulative prospect theory (CPT), and mean-variance in assessing risks. We illustrate that these models and their extensions are unable to predict risk behaviors accurately in out-of-sample empirical studies. EUT uses a nonlinear value (utility) function of consequences but is linear in probabilities, which has been criticized as its primary weakness. Although mean-variance is nonlinear in probabilities, it is symmetric, contradicts first-order stochastic dominance, and uses the same standard deviation for both risk aversion and risk proneness. In this paper, we explore a special case of geometric dispersion theory (GDT) that is simultaneously nonlinear in both consequences and probabilities. It complies with first-order stochastic dominance and is asymmetric to represent the mixed risk-averse and risk-prone behaviors of the decision makers. GDT is a triad model that uses expected value, risk-averse dispersion, and risk-prone dispersion. GDT uses only two parameters, z and zX; these constants remain the same regardless of the scale of risk problem. We compare GDT to several other risk dispersion models that are based on EUT and/or mean-variance, and identify verified risk paradoxes that contradict EUT, CPT, and mean-variance but are easily explainable by GDT. We demonstrate that GDT predicts out-of-sample empirical risk behaviors far more accurately than EUT, CPT, mean-variance, and other risk dispersion models. We also discuss the underlying assumptions, meanings, and perspectives of GDT and how it reflects risk relativity and risk triad. This paper covers basic GDT, which is a special case of general GDT of Malakooti [Malakooti (2020) Geometric dispersion theory of decision making under risk: Generalizing EUT, RDEU, & CPT with out-of-sample empirical studies. Working paper, Case Western Reserve University, Cleveland.].

Download Full-text

Machine Learning Applications in Computer Vision

Image Processing ◽

10.4018/978-1-4666-3994-2.ch045 ◽

2013 ◽

pp. 896-926

Author(s):

Mehrtash Harandi ◽

Javid Taheri ◽

Brian C. Lovell

Keyword(s):

Machine Learning ◽

Computer Vision ◽

Object Recognition ◽

Visual Recognition ◽

Sequential Decision ◽

Biologically Inspired ◽

Transductive Learning ◽

Machine Learning Applications ◽

Appearance Models ◽

Learning Frameworks

Recognizing objects based on their appearance (visual recognition) is one of the most significant abilities of many living creatures. In this study, recent advances in the area of automated object recognition are reviewed; the authors specifically look into several learning frameworks to discuss how they can be utilized in solving object recognition paradigms. This includes reinforcement learning, a biologically-inspired machine learning technique to solve sequential decision problems and transductive learning, and a framework where the learner observes query data and potentially exploits its structure for classification. The authors also discuss local and global appearance models for object recognition, as well as how similarities between objects can be learnt and evaluated.

Download Full-text

Learning to Segment Images Using Dynamic Feature Binding

Neural Computation ◽

10.1162/neco.1992.4.5.650 ◽

1992 ◽

Vol 4 (5) ◽

pp. 650-665 ◽

Cited By ~ 34

Author(s):

Michael C. Mozer ◽

Richard S. Zemel ◽

Marlene Behrmann ◽

Christopher K. I. Williams

Keyword(s):

Object Recognition ◽

Phase Locking ◽

Signal Amplitude ◽

Feature Binding ◽

Dynamic Feature ◽

Training Procedure ◽

Segmentation Process ◽

Visual Scenes ◽

Computational Systems ◽

Complex Valued

Despite the fact that complex visual scenes contain multiple, overlapping objects, people perform object recognition with ease and accuracy. One operation that facilitates recognition is an early segmentation process in which features of objects are grouped and labeled according to which object they belong. Current computational systems that perform this operation are based on predefined grouping heuristics. We describe a system called MAGIC that learns how to group features based on a set of presegmented examples. In many cases, MAGIC discovers grouping heuristics similar to those previously proposed, but it also has the capability of finding nonintuitive structural regularities in images. Grouping is performed by a relaxation network that attempts to dynamically bind related features. Features transmit a complex-valued signal (amplitude and phase) to one another; binding can thus be represented by phase locking related features. MAGIC's training procedure is a generalization of recurrent backpropagation to complex-valued units.

Download Full-text

De Verwerving Van Morfologische Regels in Schrift (III)

Toegepaste Taalwetenschap in Artikelen ◽

10.1075/ttwia.61.09dor ◽

1999 ◽

Vol 61 ◽

pp. 97-110

Author(s):

Marjolein van Dort-Slijper ◽

Gert Rijlaarsdam ◽

Eva Breedveld

Keyword(s):

Negative Transfer ◽

Empirical Studies ◽

A Priori ◽

Recognition Task ◽

Past Tense ◽

Production Task ◽

The Past ◽

Verb Inflection ◽

Group 6 ◽

Special Case

In order to provide textbook authors with empirical data on the acquisition in Dutch of written morphology in nouns, verbs and adjectives, several empirical studies have been undertaken. In this article, the third study reports on the performance of the morpheme -e in a special case of adjectives in Dutch: the adjectives derived from participles. The study tries to determine the possible interference between the morphological rules for verb inflection (past tense) and adjective declension in reading and writing. Five classes of adjectives were distinguished according to order of relative difficulty established a priori. Subjects (n=157, grade 6, 7 and 8 from two schools) individually completed a compre-hension and a production task in which factors were systematically varied. Also a recognition test on the spelling of the past tense of verbs was administered. The results showed an effect of categories of verbal adjectives in the production task, but only for groups 7 and 8; group 6 was not sensitive to the differences between the categories. In the recognition task, no effect of type of adjective (verbal or normal) was found for groups 7 and 8; but for group 6, performance on verbal adjectives was lower for the three most difficult categories of adjectives. In the production task, all three groups performed lower on verbal adjectives than normal adjectives in the two most difficult categories of adjectives. It turned out that groups which acquired spelling rules for the past tense of verbs to a higher level, made more errors in the spelling of verbal adjectives, especially in the two categories of adjectives which related the strongest to the spelling of verbs. It was concluded that indications were found that negative transfer or interference is present. Authors recommend changing the order of phases in which spelling rules are trained: from 'adjective declension-verb inflection (past tense)-verbal adjective declension' to 'adjective declension (including verbal adjective declension)-verb declension (past tense).

Download Full-text

De Verwerving van Morfologische Regels in Schrift

Toegepaste Taalwetenschap in Artikelen ◽

10.1075/ttwia.60.10dor ◽

1998 ◽

Vol 60 ◽

pp. 87-100

Author(s):

Marjolein van Dort-Slijper ◽

Gert Rijlaarsdam ◽

Maaike Ditzel

Keyword(s):

Empirical Data ◽

Cognitive Task ◽

Empirical Studies ◽

Production Task ◽

Grammatical Function ◽

Reading And Writing ◽

Stressed Syllable ◽

Special Case ◽

Grammatical Functions

In order to provide authors of text books with empirical data on the acquisition in Dutch of written morphology in nouns, verbs and adjectives, several empirical studies were undertaken. In this article, the second study is reported on the performance of the morpheme -en in a special case of adjectives in Dutch: material adjectives such as 'gouden' (gold). The study tried to determine factors which affect the performance in reading and writing. Factors involved were two interference factors (plural in adjacent nouns; normal adjectives), cognitive task (comprehension versus writing) and two syntactic factors (grammatical function and position in the sentence of the constituent. Subjects (n=80, grades 6 and 7, from four schools) individually completed a comprehension and a production task in which factors were systematically varied. No effects of cognitive task were observed. Therefore only results for the production task were reported in detail. It turned out that material adjectives were more difficult than normal adjectives, and that within the category of material adjectives two subcategories should be distinguished, the easier one in which the morpheme is preceded by a stressed syllable as in 'gouden', and the more difficult one in which this is not the case, as in 'zilveren'. Of the two syntactic factors, only the grammatical function seems to affect the performance: adjectives in constituents with subject function were more difficult than adjectives with other grammatical functions; interactions between group and category of adjectives were found. Of the two interference factors, both factors seemed to affect performance.

Download Full-text

Number detectors spontaneously emerge in a deep neural network designed for visual object recognition

Science Advances ◽

10.1126/sciadv.aav7903 ◽

2019 ◽

Vol 5 (5) ◽

pp. eaav7903 ◽

Cited By ~ 18

Author(s):

Khaled Nasr ◽

Pooja Viswanathan ◽

Andreas Nieder

Keyword(s):

Neural Network ◽

Object Recognition ◽

Visual System ◽

Deep Neural Network ◽

Number Sense ◽

Discrimination Performance ◽

Visual Object ◽

Visual Object Recognition ◽

Biologically Inspired ◽

Number Discrimination

Humans and animals have a “number sense,” an innate capability to intuitively assess the number of visual items in a set, its numerosity. This capability implies that mechanisms to extract numerosity indwell the brain’s visual system, which is primarily concerned with visual object recognition. Here, we show that network units tuned to abstract numerosity, and therefore reminiscent of real number neurons, spontaneously emerge in a biologically inspired deep neural network that was merely trained on visual object recognition. These numerosity-tuned units underlay the network’s number discrimination performance that showed all the characteristics of human and animal number discriminations as predicted by the Weber-Fechner law. These findings explain the spontaneous emergence of the number sense based on mechanisms inherent to the visual system.

Download Full-text