Neuroscience and Neurobiology Commons^™

Capturing The Representational Dynamics Of Face Perception In Deep Recurrent Neural Networks, Hossein Adeli, Chase W. King, Nikolaus Kriegeskorte

MODVIS Workshop

We investigate the representational dynamics of recurrent convolutional neural networks (RCNNs) in order to understand the time-course of visual recognition. We explore a family of models with bottom-up and lateral connections that were optimized for face-identification and object-recognition tasks. Using representational similarity analysis (RSA), we observed that only models that were trained for face identification showed a late-emerging prominent distinction of identities as seen in the monkey face patch AM. Early model responses strongly separated the objects from the faces. These findings suggest that the dynamics of face recognition that emerges in a hierarchical recurrent neural network prioritizes category-level recognition …

Attention Improves Classification Performance Of Locations In Ventral And Dorsal Visual Pathways, Jinho Lee, Sung-Mu Lee, Ishita Agarwal, Anne Sereno

MODVIS Workshop

Visual processing is divided into a ventral and dorsal pathway, thought to be important for object recognition and spatial cognition, respectively. Yet both streams encode shape and spatial information, and both are modulated by attention and task demands. This study compared how attention influences fMRI bold responses in a ventral region (fusiform gyrus, FG) and dorsal region (superior parietal gyrus, SPG). Specifically, using multivariate pattern analysis within each region, we examined whether attention modulated spatial location classification performance. During fMRI scans, participants performed one of 3 possible tasks: a 1-back task under attend-to-shape (detect a shape repetition); a 1-back task …

Comparing Apples To Oranges From Photoreceptors To Inferotemporal Cortex, Qasim Zaidi, Xu Yi, Akihito Maruya, Bevil Conway

MODVIS Workshop

No abstract provided.

Perceptograms And Cortical Models For Amblyopic Form Distortions, Akihito Maruya, Farzaneh Olianezhad, Jingyun Wang, Jose-Manuel Alonso, Qasim Zaidi

MODVIS Workshop

No abstract provided.

Influence Of Task On The Geometry Of A Perceptual Space, Jonathan D. Victor, Mary M. Conte

MODVIS Workshop

Perceptual spaces are representations of a sensory or cognitive domain in which the domain’s elements correspond to points, and distances between these points are perceptual differences. Proximity relationships within a perceptual space can support a variety of functions, including discrimination, grouping, learning, and generalization. These diverse functions may use the features of the domain in different ways, resulting in task-dependent influences on the geometry of the space.

To identify and characterize these influences, we focused on a domain of visual textures. These textures varied across many dimensions, including mean luminance and low- and high-order spatial correlations, which formed the axes …

Using Neural Networks To Better Understand Static And Dynamic Components Of Facial Expression Recognition, Yi-Fan Li, Anne Bibiana Sereno

MODVIS Workshop

Facial expressions are crucial social information for human communication. In the real world, facial expressions are dynamic; however, much of existing research in facial expressions relies on static stimuli. This static approach may limit the ecological validity of our understanding of the emotional information on faces. Our study aims to investigate the dynamic and static information contained in different emotional categories of facial expressions (e.g., happy, sad). Four convolutional neural networks are introduced as models to be trained with a large-scale dynamic facial expression dataset (short videoclips of 16 frames of 7 different emotional categories) in four different ways: ordered …

Modeling Effects Of Edge Classification And Perceptual Organization On The Appearance Of Disk/Annulus Stimuli, Michael E. Rudd

MODVIS Workshop

In classical theories of sensory psychophysics, appearance matches conducted with simple stimuli consisting of disks surrounded by annuli were posited to probe low-level contrast mechanisms in the visual pathways that computed the approximate luminance ratio between the disk and annulus for the purpose of achieving lightness constancy (Wallach, 1948). Subsequent models have explained such matches on the basis of neural edge integration (Shapley & Reid, 1985; Rudd & Zemach, 2004). Here, I demonstrate that appearances matches made in the disk/annulus paradigm are influenced by at least two types of top-down effects involving edge classification and perceptual organization. I present a …

Fixational Eye Movements Shake Up The Stationary View On Pattern Vision, Lynn Schmittwilken, Marianne Maertens

MODVIS Workshop

No abstract provided.

A Computational Perspective On Segregated Processing And The Privileged Role Of Space In Binding, Anne Sereno, Zhixian Han, Elizabeth M. Frazier

MODVIS Workshop

No abstract provided.

Fusing Classic Motion Energy Models And Deep Learning For Coarse-To-Fine Moving Object Segmentation, Matthias Tangemann, Matthias Kümmerer, Matthias Bethge

MODVIS Workshop

Classic motion energy models are able to predict a wide range of physiological and behavioral aspects of motion perception in humans. Whether these models can be used as a basis for higher-level tasks, such as moving object segmentation, has however hardly been explored yet. Here, we present a model that combines a motion energy representation with recent computer vision approaches for figure-ground segmentation of naturalistic stimuli. We find that unlike established motion segmentation models but similar to humans, our model generalizes to random-dot stimuli when only trained on RGB videos.

Perceptual Grouping With Latent Noise, Ben Lonnqvist, Zhengqing Wu, Michael H. Herzog

MODVIS Workshop

Humans effortlessly group elements into objects and segment them from the background and other objects without supervision. For example, the black and white stripes of a zebra are grouped together despite vastly different colors. A thorough theoretical and empirical account of perceptual grouping is still missing – Deep Neural Networks (DNNs), which are considered leading models of the visual system still regularly fail at simplistic perceptual grouping tasks. Here, we propose a counterintuitive unsupervised computational approach to perceptual grouping and segmentation: that they arise because of neural noise, rather than in spite of it. We show that adding noise in …

Facilitation In Pattern Motion Perception Of Self-Operated Stimuli Explained By Adaptive Contrast Normalization, Fulvio Missoni, Francesca Peveri, Andrea Canessa, Giulia Sedda, Vittorio Sanguineti, Silvio P. Sabatini

MODVIS Workshop

Movement can affect the way we make sense of complex visual information. To investigate this issue, we designed an experiment to assess changes of plaid motion perception threshold after a period of sensorimotor contingency experience. We found that movement training facilitates combination of elementary motion cues into a global motion percept. No changes in perceptual thresholds are observed in a passive visual condition. A Bayesian model suggested a reduction, after training, of the cross-talk between two gratings with unbalanced contrasts in corresponding sensory channels. To test plausible neural mechanisms for active reduction of cross-talk in cortical representation of complex visual …

Normalization With Dynamic Weights Predicts Neural And Behavioral Effects Of Orientation Adaptation, Huseyin Boyaci

MODVIS Workshop

Prolonged exposure to an oriented contour causes adaptation and has nontrivial effects on neural activity and perception. For example, the neuron's response amplitude may change (suppression or facilitation), the width of its tuning curve may change (broadening or sharpening), and its preferred orientation may shift (repulsion or attraction). Perceptually, adaptation affects the perceived orientation of a subsequently presented contour (direct and indirect tilt aftereffect) and alters orientation discrimination thresholds. In this study, I show that the normalization model with dynamic weights can predict these empirical results.

Local Geometry Of Elementary Visual Computations, Peter Neri

MODVIS Workshop

Visual operators (e.g. edge detectors) are classically modelled using small circuits involving canonical computations, such as template-matching and gain control. Circuit models explain many aspects of the empirical descriptors that are used to characterize local visual operators, from sensitivity to classification images. Notwithstanding their utility, these models fail to provide a unified framework encompassing the variety of effects observed experimentally, such as the impact of contrast, SNR, and attention on the above descriptors. My goal is to start with a simple, plausible geometrical representation of the perceptual operation carried out by the observer, and to show that this representation is …

Do Mechanisms Of Sinusoidal Contrast Sensitivity Account For Edge Sensitivity?, Lynn Schmittwilken, Felix A. Wichmann, Marianne Maertens

MODVIS Workshop

No abstract provided.

Characterization Of Perceptual Spaces From Triadic Similarity Judgments, Jonathan Victor

MODVIS Workshop

Perceptual spaces are mental workspaces that organize a sensory or cognitive domain into a format that supports functions such as comparison, grouping, learning, and generalization. Determining the geometry of a perceptual space – i.e., the properties of the perceptual distances within the space -- is thus crucial not only to understand these intermediate levels of processing at an algorithmic level, but also as a starting point for comparison with neural measures of similarity. While perceptual spaces are often taken to be Euclidean or nearly so (the classic example is trichromatic color space), this assumption, when tested, is often violated. Moreover, …

A Foveated Model Of Visual Discrimination Based On Windowed Texture Statistics, Jan W. Kurzawski, William F. Broderick, Eero P. Simoncelli, Jonathan Winawer

MODVIS Workshop

Most information from visual scenes is discarded by the human nervous system and thus cannot influence visual behavior. Here, we investigated the loss of spatial pattern information. This loss increases dramatically with eccentricity. Recently, models have been developed that average image features in pooling windows whose diameters scale with eccentricity. These models can be used to synthesize “metamers” of natural scenes, images which are physically different but perceptually indistinguishable. Psychophysical experiments have identified the maximum window scaling for which model discrimination abilities match human performance (“critical scaling”). If images are synthesized with pooling windows that exceed critical scaling (“supercritical scaling”), …

Dissecting Bayes: Using Influence Measures To Test Normative Use Of Probability Density Information Derived From A Sample, Laurence T. Maloney, Keiji Ota

MODVIS Workshop

Bayesian decision theory (BDT) is used to model human performance in tasks where the decision maker must compensate for uncertainty in order to gain rewards and avoid losses. BDT prescribes how the decision maker can combine available data, prior knowledge, and value to reach a decision maximizing expected winnings. Do human decision makers actually use BDT in making decisions? Researchers typically compare overall human performance (total winnings or overall percent correct) to the predictions of BDT but we cannot conclude that BDT is an adequate model for human performance based on just overall performance. We break BDT down into elementary …

Explaining The Staircase Gelb Illusion, Simultaneous Contrast, And Perceptual Fading Of Stabilized Images With A Neural Model Driven By Fixational Eye Movements, Michael E. Rudd

MODVIS Workshop

A neural model of lightness computation driven by fixational eye movements is described and used to simulate various lightness phenomenon, including the Staircase Gelb illusion and its variants, simultaneous contrast, the Chevreul illusion, and perceptual fading of stabilized images. The model provides a precise account of the lightness matches from several experiments, with an overall error of only 1.5%. In the model, spatial maps of transient ON and OFF cell activations—produced as the eyes traverse the visual scene—are sorted by eye movement direction in visual cortex. At a subsequent processing stage, the activations within these maps are summed across space …

Physiological Rationale For Fixation Eye-Movements, Qasim Zaidi

MODVIS Workshop

No abstract provided.

Task-Driven Influences On Fixational Eye Movements, Jonathan Victor, Yen-Chu Lin, Michele Rucci

MODVIS Workshop

There is now compelling evidence that the spatiotemporal remapping carried out by fixational eye movements (FEMs) is an essential step in visual processing. Moreover, the overall Brownian-like statistics of FEMs are calibrated to map fine spatial detail into the temporal frequency range to which retinal circuitry is tuned. Here, we tested the hypothesis that the detailed spatial characteristics of FEMs can be adjusted to task demands via cognitive influences that operate even in the absence of a visual stimulus. We examined FEMs in a task that required subjects (N=6) to report which of two letters was displayed. Trials were blocked; …

Active Encoding Of Space Through Time, Michele Rucci, Jonathan D. Victor

MODVIS Workshop

No abstract provided.

Extracting Edges In Space And Time During Visual Fixations, Lynn Schmittwilken, Marianne Maertens

MODVIS Workshop

No abstract provided.

Automated Delineation Of Visual Area Boundaries And Eccentricities By A Cnn Using Functional, Anatomical, And Diffusion-Weighted Mri Data, Noah C. Benson, Bogeng Song, Toshikazu Miyata, Hiromasa Takemura, Jonathan Winawer

MODVIS Workshop

Delineating visual field maps and iso-eccentricities from fMRI data is an important but time-consuming task for many neuroimaging studies on the human visual cortex because the traditional methods of doing so using retinotopic mapping experiments require substantial expertise as well as scanner, computer, and human time. Automated methods based on gray-matter anatomy or a combination of anatomy and functional mapping can reduce these requirements but are less accurate than experts. Convolutional Neural Networks (CNNs) are powerful tools for automated medical image segmentation. We hypothesize that CNNs can define visual area boundaries with high accuracy. We trained U-Net CNNs with ResNet18 …

Toward A Manifold Encoding Neural Responses, Luciano Dyballa, Andra M. Rudzite, Mahmood S. Hoseini, Mishek Thapa, Michael P. Stryker, Greg D. Field, Steven W. Zucker

MODVIS Workshop

Understanding circuit properties from physiological data presents two challenges: (i) recordings do not reveal connectivity, and (ii) stimuli only exercise circuits to a limited extent. We address these challenges for the mouse visual system with a novel neural manifold obtained using unsupervised algorithms. Each point in our manifold is a neuron; nearby neurons respond similarly in time to similar parts of a stimulus ensemble. This ensemble includes drifting gratings and flows, i.e., patterns resembling what a mouse would “see” running through fields.

Regarding (i), our manifold differs from the standard practice in computational neuroscience: embedding trials in neural coordinates. Topology …

Constraining The Binding Problem Using Maps, Zhixian Han, Anne Sereno

MODVIS Workshop

We constrained the binding problem by creating maps of different attributes. We compared the performance of different models with different maps in our current study. Our preliminary results showed that the performance of the model is the highest when location maps were used. These results suggest that the optimal way to constrain the binding problem is to create location maps of different attributes.

From Image Gradients To A Perceptual Metric Space, Alan Johnston

MODVIS Workshop

How do we achieve a sense of spatial dimension from a sense of location? There are three predominant ideas about how we achieve this; spatial isomorphism, in which what we see reflects differences in distance or size in the brain; that spatial extent depends upon motor sensations or intentions related to eye movements; and that distance is computed from the degree of correlation in neural activity between adjacent locations, with distance inversely proportional to the correlation. There are problems with each of these approaches, for example, neural correlation may depend more on image structure than adjacency - consider the case …

V1 Saliency Hypothesis And Central-Peripheral Dichotomy (Cpd), Li Zhaoping Prof. Dr.

MODVIS Workshop

No abstract provided.

A Dynamical Model Of Binding In Visual Cortex During Incremental Grouping And Search, Daniel Schmid, Daniel A. Braun, Heiko Neumann

MODVIS Workshop

Binding of visual information is crucial for several perceptual tasks. To incrementally group an object, elements in a space-feature neighborhood need to be bound together starting from an attended location (Roelfsema, TICS, 2005). To perform visual search, candidate locations and cued features must be evaluated conjunctively to retrieve a target (Treisman&Gormican, Psychol Rev, 1988). Despite different requirements on binding, both tasks are solved by the same neural substrate. In a model of perceptual decision-making, we give a mechanistic explanation for how this can be achieved. The architecture consists of a visual cortex module and a higher-order thalamic module. While the …

Efficient Coding Of Local 2d Shape, James Elder, Timothy D. Oleskiw, Ingo Fruend, Gerick M. Lee, Andrew Sutter, Anitha Pasupathy, Eero Simoncelli, J Anthony Movshon, Lynne Kiorpes, Najib Majaj

MODVIS Workshop

Efficient coding provides a concise account of key early visual properties, but can it explain higher-level visual function such as shape perception? If curvature is a key primitive of local shape representation, efficient shape coding predicts that sensitivity of visual neurons should be determined by naturally-occurring curvature statistics, which follow a scale-invariant power-law distribution. To assess visual sensitivity to these power-law statistics, we developed a novel family of synthetic maximum-entropy shape stimuli that progressively match the local curvature statistics of natural shapes, but lack global structure. We find that humans can reliably identify natural shapes based on 4^th and …