Browsing by Subject "Vision"

Now showing 1 - 15 of 15

Context-dependent adaptation in the visual system
(2017-12) Mesik, Juraj
The visual system continuously adjusts its sensitivities to various visual features so as to optimize neural processing, a phenomenon known as adaptation. Although this rapid form of plasticity has been extensively studied across numerous sensory modalities, it remains unclear if its dynamics can change with experience. Specifically, the world we live in is composed of many different environments, or contexts, each of which contains its own statistical regularities. For example, forests contain more vertical energy and greenish hues than a desert landscape. Here we investigated the possibility that through experience, the visual system can learn statistical regularities in the visual input, and use this knowledge to adapt more quickly. In two sets of experiments, participants repeatedly adapted to previously unexperienced regularities in orientation statistics over the course of 3-4 sessions. They adapted either to rapidly presented sequences of oriented gratings containing orientation biases, or to natural visual input that was filtered to alter its orientation statistics. We found that experience did increase adaptation rate, but only in the experiments where participants adapted to a single set of altered statistics of natural input. We found no changes in adaptation rate in experiments where participants periodically switched between adapting to different statistical regularities. These results demonstrate that adaptation and experience can interact under some circumstances.
Data for Validating a Model of Architectural Hazard Visibility with Low-Vision Observers
(2020-07-22) Liu, Siyun; Thompson, William B.; Liu, Yichen; Shakespeare, Robert A.; Kersten, Daniel J.; Legge, Gordon E.; liux4433@umn.edu; Liu, Siyun; Department of Psychology, University of Minnesota; School of Computing, University of Utah; Department of Theatre, Drama, and Contemporary Dance, Indiana University Bloomington
Pedestrians with low vision are at risk of injury when hazards, such as steps and posts, have low visibility. This study aims at validating the software implementation of a computational model that estimates hazard visibility. The model takes as input a photorealistic 3-D rendering of an architectural space, and the acuity and contrast sensitivity of a low-vision observer, and outputs estimates of the visibility of hazards in the space. Our experiments explored whether the model can predict the likelihood of observers correctly identifying hazards. We tested fourteen normally sighted subjects with blur goggles that reduced acuity to 1.2 logMAR or 1.6 logMAR and ten low-vision subjects with acuities ranging from 0.8 logMAR to 1.6 logMAR. Subjects viewed computer-generated images of a walkway containing five possible targets ahead—large step up, large step-down, small step up, small step down, or a flat continuation. Each subject saw these stimuli with variations of lighting and viewpoint in 250 trials and indicated which of the five targets was present. The model generated a score on each trial that estimated the visibility of the target. If the model is valid, the scores should be predictive of how accurately the subjects identified the targets. We used logistic regression to examine the correlation between the scores and the participants’ responses. For twelve of the fourteen normally sighted subjects with artificial acuity reduction and all ten low-vision subjects, there was a significant relationship between the scores and the participant’s probability of correct identification. These experiments provide evidence for the validity of a computational model that predicts the visibility of architectural hazards. The software implementation of the model may be useful for architects to assess the visibility of hazards in their designs, thereby enhancing the accessibility of spaces for people with low vision.
Foraging mechanisms of siscowet Lake Trout (Salvelinus namaycush siscowet) on pelagic prey
(2014-12) Keyler, Trevor Daniel
Little is known about the foraging mechanisms of siscowet (Salvelinus namaycush siscowet), the most abundant piscivore within Lake Superior because they occupy light-limited environments as a result of diel vertical migration. The reaction distance, angle of attack, and foraging success were determined for siscowet during laboratory trials under lighting conditions that approximated downwelling spectral irradiance and intensity (0-10 lx) at daytime depths. Siscowet reaction distance in response to golden shiners (Notemigonus crysoleucas) was directly correlated with increasing light intensity until saturation at 0.01 lx, and afterwards the relationship was asymptotic within our range of tested light intensities. In total darkness, lateral line sensory detection was sufficient to locate prey at 24.9 ± 1.7 cm, while increasing light intensities increased reaction distance up to 58.6 ± 2.3 cm at 10 lx. Larger prey elicited higher reaction distances than smaller prey at all light intensities while moving prey elicited higher reaction distances than stationary prey at the higher light intensities (0.001-10 lx). The capture and consumption of prey similarly increased with increasing light intensity while time to capture decreased with increasing light intensity. The majority of orientations toward prey occurred within 120° of the longitudinal axis of the siscowet's eyes, although reaction distances among 30° increments along the axis were not significantly different. Our predictive model will help determine reaction distances for siscowet in various photic environments and will help identify the mechanisms and behavior that allow for low light intensity foraging within freshwater systems.t>
Freeway Network Traffic Detection and Monitoring Incidents
(Minnesota Department of Transportation, 2007-10) Joshi, Ajay J.; Atev, Stefan; Fehr, Duc; Drenner, Andrew; Bodor, Robert; Masoud, Osama; Papanikolopoulos, Nikolaos P.
We propose methods to distinguish between moving cast shadows and moving foreground objects in video sequences. Shadow detection is an important part of any surveillance system as it makes object shape recovery possible, as well as improves accuracy of other statistics collection systems. As most such systems assume video frames without shadows, shadows must be dealt with beforehand. We propose a multi-level shadow identification scheme that is generally applicable without restrictions on the number of light sources, illumination conditions, surface orientations, and object sizes. In the first level, we use a background segmentation technique to identify foreground regions that include moving shadows. In the second step, pixel-based decisions are made by comparing the current frame with the background model to distinguish between shadows and actual foreground. In the third step, this result is improved using blob-level reasoning that works on geometric constraints of identified shadow and foreground blobs. Results on various sequences under different illumination conditions show the success of the proposed approach. Second, we propose methods for physical placement of cameras in a site so as to make the most of the number of cameras available.
Highly Selective Attentional Modulation of Task-Appropriate Neural Populations in Primary Visual Cortex
(2017-06) Warren, Scott
A wide variety of different forms of attention have been described in the human and non-human literature, however the recently developed Input Gain Model of visual attention proposes that a simple neural mechanism, multiplicative gain, may be employed to explain much of the available data on visual attentional modulations. On this basis, we hypothesized that a better explanation for distinct forms of attention may be that this simple attentional mechanism is in fact highly specific: attentional modulations are only present within task-appropriate neurons or neuron groups, and it is the location (and not nature) of these modulations which defines the observer’s current attentive state. We present the results of two orthogonal attention tasks, both targeting distinct but specific and well defined sub-populations of primary visual cortical (V1) neurons. In both experiments we observe that attentional modulations are grossly targeted to neural populations that are selectively tuned for the cue. When humans attend to one orientation, voxels reflecting orientation selective neurons tuned toward that orientation are selectively enhanced. When monkeys were trained to attend to a very small region of space, attention modulated the V1 representation of stimulus elements near that location in space. In both studies, these modulations are predictive of observer behavior, providing evidence that attentional modulation of V1 meaningfully impacts the perceptibility of the attended stimuli. Systematic imprecision in these modulations suggest that attentional modulations of V1 are mediated through corticocortical feedback, hypothetically from secondary visual cortex. This provides a strong constraint for further refinement of general models of attention.
Indoor Spatial Updating with Reduced Visual Information
(2015-09-15) Legge, Gordon, E; Gage, Rachel, J; Baek, Yihwa; Bochsler, Tiana, M; legge@umn.edu; Legge, Gordon, E
Spatial updating refers to the ability to keep track of position and orientation while moving through an environment. People with impaired vision may be less accurate in spatial updating with adverse consequences for indoor navigation. In this study, we asked how artificial restrictions on visual acuity and field size affect spatial updating, and also judgments of the size of rooms. Normally sighted young adults were tested with artificial restriction of acuity in Mild Blur (Snellen 20/135) and Severe Blur (Snellen 20/900 ) conditions, and a Narrow Field (8º) condition. The subjects estimated the dimensions of seven rectangular rooms with and without these visual restrictions. They were also guided along three-segment paths in each of the rooms. At the end of each path, they were asked to estimate the distance and direction to the starting location. In Experiment 1, the subjects walked along the path. In Experiment 2, they were pushed in a wheelchair to determine if reduced proprioceptive input would result in poorer spatial updating. With unrestricted vision, mean Weber fractions for room-size estimates were near 20%. Severe Blur but not Mild Blur yielded larger errors in room-size judgments. The Narrow Field was associated with increased error, but less than with Severe Blur. There was no effect of visual restriction on estimates of distance back to the starting location, and only Severe Blur yielded larger errors in the direction estimates. Contrary to expectation, the wheelchair subjects did not exhibit poorer updating performance than the walking subjects, nor did they show greater dependence on visual condition. If our results generalize to people with low vision, severe deficits in acuity or field will adversely affect the ability to judge the size of indoor spaces, but updating of position and orientation may be less affected by visual impairment.
Investigating rapid divergence of sensory systems between satellite populations of the Mexican tetra (Astyanax mexicanus)
(2022-01) Enriquez, Maya
The Mexican tetra (Astyanax mexicanus) has two primary ecotypes: cave fish and surface fish. Cave fish are characterized by troglomorphic phenotypes, such as vestigial eyes and reduced pigmentation. Studies have documented phenotypic differences in these ecotypes, which likely diverged between 0.2 to 1 million years ago. However, surface A.mexicanus fish were introduced relatively recently to the Edwards-Trinity aquifer in Texas in the early 1900s, and subsequent cave colonization by portions of this population show evidence of divergence through rapid phenotypic and behavioral evolution. The establishment of these satellite populations from cave and surface river invasions are a case study into the rapid evolution of traits within a new environment, allowing observation on how sensory systems may adapt in real time. Auditory evoked potentials (AEPs), particle acceleration (PAL) and electroretinography (ERG) assays were conducted to quantify sensory differences between satellite cave and surface populations. Honey Creek cave fish were found to be significantly more sensitive (p < 0.05) than Honey Creek surface fish to sound pressure levels between 0.5 kHz - 0.8 kHz, while some pairwise differences were found between San Antonio Zoo surface, Blue Hole cave and San Pedro Springs cave populations between 0.5 kHz - 0.7kHz (p < 0.05). Particle acceleration assays also showed significant differences between Honey Creek cave and surface (p < 0.05) as well as San Antonio Zoo surface, San Pedro Springs cave and Blue Hole cave (p < 0.05) within the same range of frequencies tested. Electroretinography data indicated that Honey Creek cave fish were significantly less sensitive (p < 0.05) to light than Honey Creek surface fish at 530 nm, while no differences were found between San Antonio Zoo surface, San Pedro Springs cave and Blue Hole cave. Collectively, these results indicate rapid divergence of A.mexicanus in cave habitats at the most sensitive ranges of their visual and auditory sensory systems, and future monitoring may demonstrate continual divergence of sensory systems in populations exposed to new environments.
Local and Iterative Visual Processing Deficits in Schizophrenia
(2017-05) Espensen-Sturges, Tori
Evidence of dysfunctional visual processing in schizophrenia patients has been noted in all stages of the visual processing pathway. The iterative nature of vision- with hierarchical feedforward signals, modulating feedback signals, and horizontal intracortical connections- makes it difficult to pinpoint exact loci that are driving these deficits. This dissertation uses several contextual modulation paradigms in an effort to isolate and explain the nature of disruptions in iterative visual processing in schizophrenia. Chapter 1 provides an overview of visual processing dysfunctions in schizophrenia, and examines a variety of mechanisms that may play a role. These include neurotransmitters, magnocellular versus parvocellular processing streams, abnormal local connections, and abnormal long-range feedback connections. These are presented in the context of several theoretical perspectives of visual neuroscience. Chapter 2 provides functional magnetic resonance imaging data from a fractured ambiguous object task that probes the role of high-level qualities in primary visual cortex activation and interregional connectivity, and how these may be disrupted in psychosis. Chapter 3 introduces a computational model that attempts to fit parameters to psychophysical data to isolate disrupted mechanisms in schizophrenia. This model focuses on the role of gain control and segmentation of center and surround stimuli in a tilt illusion paradigm. Chapter 4 presents previous work, examining the modulatory effect of the NMDA receptor agonist d-cycoserine on conditioned fear generalization. This work was done in healthy controls as a step in expanding our knowledge of the function of d-cycloserine in increasing specificity and efficacy in the fear learning process.
More than a feeling: emotional value alters visual perception.
(2012-05) McMenamin, Brenton W.
Objects with emotional or motivational value are thought to benefit from facilitated perceptual processing. However, previous studies that have combined emotional variables with perceptual tests generally have relied on stimuli with well-established emotional value (e.g., expressive and neutral faces), ensuring that the manipulation of emotional value was accompanied by a change in perceptual content. This dissertation contains a set of experiments that combine an aversive conditioning paradigm with tests of perceptual detection and discrimination, so the perceptual characteristics of the emotional and non-emotional stimuli can be controlled and counterbalanced. Moreover, the stimulus space is designed to allow the effects of emotional value and emotional ambiguity to be deconfounded. Behavioral results indicate that perceptual detection and discrimination abilities are improved for stimuli with emotional value, however the effect has a delayed onset and only appears in the second half of the experiment. Emotional value is associated with a decreased amplitude for early visually-evoked ERPs (P1 window), and sLORETA source-estimation suggests that this P1 amplitude reduction is caused by increased activity in orbitofrontal cortex and ventromedial PFC. The P1 effect occurs in the first half of the experiment, but not the second, suggesting that it may be a reinforcement learning signal used to drive perceptual reorganization. The final experiment explores whether conditioning-induced changes to target detection would generalize to novel stimuli and how the categorical structure of the stimulus space can impact the effects of emotion.
Neural mechanisms of visual context processing in healthy adults and those with Schizophrenia
(2014-12) Schallmo, Michael-Paul
The brain's response to a visual stimulus depends in part on the context in which it appears. For example, objects appearing within similar-looking backgrounds tend to evoke smaller neural responses than those seen in isolation. While it is known that schizophrenia (SZ) may reduce visual context effects, the neural mechanisms involved are not fully understood. This dissertation uses functional magnetic resonance imaging (fMRI) and visual behavioral tasks to examine the role of context during normal visual processing, and how context processing is affected by SZ.Chapter 1 provides an overview of the forms of contextual modulation that will be addressed later, and their impairment in SZ. Chapter 2 describes a series of five experiments probing how factors such as stimulus geometry, presentation timing, and attention affect the fMRI response to small groups of visual stimuli. In primary visual cortex, the relative strength of contextual modulation was found to increase when subjects directed their attention away from the stimuli. Further, fMRI responses to parallel center and surrounding stimuli did not show the predicted sensitivity to center contrast.In Chapters 3 and 4, the effect of spatial context during early visual processing in SZ patients was assessed using behavioral measures. Surround suppression of perceived contrast was examined in Chapter 3 among SZ patients and their unaffected relatives, as well as subjects with bipolar affective disorder (BP), relatives of BP subjects, and healthy controls. Weaker surround suppression was observed in SZ versus control subjects, while BP patients showed an intermediate deficit. These deficits did not depend on the configuration of surrounding stimuli. Normal performance was observed among relatives of SZ and BP subjects, indicating deficits in surround suppression were not associated with a genetic risk for these disorders. Chapter 4 examined how SZ impairs the ability to detect visual contours in cluttered backgrounds. Contours were presented in more- or less-similar backgrounds, in order to assess contextual modulation. While SZ patients performed worse than healthy controls or SZ relatives when detecting contours, performance in SZ was less influenced by background context. These experiments were designed to explore the neural basis of visual context processing in healthy adults, and to help uncover how SZ impairs these processes. The large body of research into the neurophysiology of human vision provides powerful tools with which to study how SZ may disrupt neural processing. Studying visual context processing may ultimately help to uncover computational principles conserved across many neural systems, and aid in identifying new targets for the treatment of mental disorders.
Predicting face recognition skills in children: global processing and attention to the eyes
(2013-05) Corrow, Sherryse
Faces are, arguably, the most important stimulus in our lives. Yet, we understand very little about what information is used to recognize faces. Two theories exist in the literature on this topic. First, it is widely believed that successful face recognition depends on the ability to utilize configural, or holistic, information about the face. Second, many have speculated that attention to the eye region of the face is essential for successful face recognition. However, few studies, none with children, have directly evaluated this relationship by examining individual differences in face processing. Thus, the goal of the following studies was to examine how individual differences in face recognition skills are predicted by configural processing of the face and, in particular, attention to the eye region. Across four experiments, children completed face recognition tasks using an eye tracker, tasks of configural processing, and an object recognition task. Results from Experiments 1 and 2 support the notion that attention to the eye region and configural processing of faces as measured by the Part-Whole Task are predictive of face recognition scores as measured by the Cambridge Face Memory Task for Children. Furthermore, these experiments provide preliminary evidence that attention to the outer areas of the face, such as the forehead, may inhibit face recognition ability. Experiment 3 generally replicated these findings, with a few exceptions, by examining a pre-selected group of children and subsequently comparing high and low performers on the Cambridge Face Memory Task. Finally, Experiment 4 examined six children with developmental prosopagnosia. Results from this experiment suggest that children with prosopagnosia are a very heterogeneous group. The results of these studies generally support three hypotheses: 1. Children who demonstrate greater attention to the eye region perform better on tasks of face recognition, 2. Higher scores on tests of configural face processing predict higher scores on tasks of face recognition, and 3. Children who demonstrate a greater degree of configural face processing are more likely to attend to the eye region of the face (Experiment 2).
Representation of Human Body Stimuli within the Human Visual System
(2021-03) Bratch, Alexander
The human body is a unique and complex visual stimulus, the accurate representation of which is critical for social interaction/communication, acquisition of complex skills, and even basic survival. In a series of studies, using behavioral and neuroimaging techniques, the visual representation of some of the basic features of body perception (e.g., individual parts and their relationships) were explored. First, using a behavioral adaptation paradigm, sensitivity to the relative proportions between limbs was explored. We found that human observers were highly sensitive to relative limb proportion and further demonstrated that this effect appears to depend on body and limb specific mechanisms. Second, an fMRI experiment was used to assess the sensitivity of body-selective cortical areas to the spatial configuration of pairs of limb parts. We found that activity in body-selective areas systematically varies with the typicality of their spatial configuration. Finally, advanced sub-millimeter fMRI techniques were used investigate whether body-selective cortex contained subordinate representations of body stimuli (e.g., their individual parts). We found that the body-selective area of the right extrastriate cortex yields the highest responses to hands above all other stimuli. Furthermore, advanced spatial mapping techniques revealed that this cortical area contains spatially consistent clusters of voxels which preferentially respond to hands. Taken together, these results help to clarify how individual body parts and the relationships between them are represented within the human visual system.
Temporal and spatial properties of the BOLD fMRI response to first and second order contrast in V1.
(2010-01) Thompson, Serena Kainoa Au
Our ability to detect, discriminate, and identify objects, to extract depth information through the use of stereopsis, and to learn about and classify different surface properties and textures are all secondary to first extracting useful cues about the local contrast within a visual scene. Contrast can be made by a change in a variety of visual features. We will specifically consider the contrast made by first-order and second-order luminance changes, specifically the contrast between local pixel luminance, and the contrast between local pixel luminance variance. There are a variety of tools available to `observe' cortical modulation to contrast, such as electrophysiological recordings of spike rate or local field potentials, fMRI BOLD modulation, magnetic encephalography (MEG), or visually evoked potentials (VEP), or through behavioral markers such as detection thresholds and discrimination thresholds. BOLD fMRI modulation provides a unique tool to measure the early visual response to local changes in image contrast with superior spatial specificity and minimal subject invasiveness, while psychophysics allows us to quantify the information human observers use to detect contrast. We will use both BOLD fMRI and psychophysics to explore three components of the human response to contrast: 1) how much do we use contrast when it forms the edge of a target shape compared to when it composes the interior area of the shape; 2) how well does second-order contrast modulate early visual cortex compared to the modulation elicited by first-order contrast; and 3) how accurate is the timing information of functional magnetic resonance imaging (fMRI) of the blood oxygen-level dependent (BOLD) modulation. Human observers do not always use contrast in the same way, and in the first experiment we study how human observers use first-order contrast in an image region versus second-order contrast at a region boundary when accomplishing a difficult detection task. We may expect V1 response to differ when contextual cues suggest that contrast is contained within an object border compared to when it fills in the interior texture of an object. Consider the case of two Holstein (black and white spotted) cows standing together, one partially occluding the other. We can use contrast cues to both identify which parts belong to which cow, as well as to determine the type of cow based on the spotted character of its hide. Both the segmentation task and the identification task require use of the same type of contrast information. However, optimal processing may change the way that low-level cues are handled in early visual cortex by modulation through context-dependent feedback as well learned information about cows. This may give rise to similar low-level cues producing different neural signals in early cortical areas, and subsequently, different sensitivity to cues in a border compared to a region. Our first experiment explores observers' use of contrast contained within the edge of a detectable target compared to the contrast that makes up the entire interior region of the target. Extraction of local image contrasts occurs at an early stage of visual processing, however, which types of contrast and how strongly they modulate early visual cortex remains undetermined. There are a variety of contrasts that could be compared. The first-order versus second-order contrast comparison is appealing because, with images generated by locally changing either average pixel luminance or pixel variance in white noise, we can create stimuli that are equal in orientation and spatial frequency information, contain detectable boundaries, and still require independent information to be extracted for detection. Our second experiment quantifies the response in primary visual cortex (V1) to first- and second-order using fMRI to acquire a spatially specific edge response in human subjects. Finally, it would be ideal to acquire both spatially and temporally specific measurements of cortical modulation in response to contrast changes. Several research groups around the world have begun using the relative timing between BOLD fMRI events to assign causal significance to the temporal order of modulation across different cortical areas using dynamic causal modeling. If, however, the timing of the modulation depends on the strength of the modulation, we could draw false conclusions from relative timing comparisons. Therefore, in our third experiment we measured three temporal characteristics of the BOLD hemodynamic response function (HRF) as a function of stimulus contrast: onset latency, time to peak, and full-width half-maximum. These explorations into the human and BOLD fMRI response to contrast are aimed at developing the current knowledge about how we perceive and parse the world around us, as well as how we can better interpret one of our measurement tools, fMRI, as a correlate of the neuronal activity occurring in early visual cortex. Both aims guide us toward a better understanding of the mechanisms we employ to process the rich information our visual systems acquire.
Vision Guidance System for the Robotic Roadway Message Painter
(2015-07) Ahmad, Waqas
Roads are used as a way of transportation and communication in every part of the world. Road markers are painted on the roads for safety, information and caution purposes. Road markers are mainly used to assist drivers with vital information and caution. The process of painting new and existing roadway markings (turn arrows, STOP messages, railroad crossings, etc.) is an important maintenance task. The Minnesota Department of Transportation (MN/DOT) estimates that over 75% of symbol and message painting is the repainting of existing markings [1]. It would be extremely valuable for an automated painting system to have a vision guidance capability whereby an existing marker could be repainted accurately with little operator input. A vision system capable of identifying existing painted pavement markings and determining their coordinates, dimensions, location, and orientation would be a very useful tool. The information regarding the pavement markings could then be passed to a robotic painting device (currently under development) to enable it to accurately repaint the marking. This would significantly improve the capability of the device to repaint existing pavement markings. Eventual users of a device using this technology could be city, county, state, federal government agencies and private companies or contractors. It will allow improved safety, reduced cost and less time to maintain existing road markers as well as draw new ones.
Visual Sensitivity, Behavior, and Habitat of Select North American Fishes
(2018-05) Keyler, Trevor
This dissertation examines how both physiological and behavioral techniques can be used to address the visual capabilities of three low-light foraging species, the walleye (Sander vitreus), siscowet lake trout (Salvelinus namaycush siscowet), and deepwater sculpin (Myoxocephalus thompsonii). Visual physiology studies can reveal the specific wavelengths fish have adapted to detect at low light intensities while behavioral studies determine the minimal light intensities needed to forage, which may ultimately reveal habitat characteristics important to individual species. Our first study used electroretinography to determine the scotopic spectral sensitivity of the walleye dark-adapted retina (peak sensitivity 500-550 nm) in addition to the approximate maximum depths where visually mediated behavior may occur during the day (77.5 m) and at night (11.3 m) for kPAR = 0.3. For our second study, we found that siscowet lake trout reaction distance to deepwater sculpin increased with increasing light intensity (up to 6.0×109 photons m-2 s-1, thereafter remaining constant), but was not affected by substrate type. Third, we determined that the average number of deepwater sculpin movements per trial increased with decreasing light intensity in the presence of siscowet lake trout, where both activity and reaction distance were suppressed at upper light intensities. Finally, we used solar/lunar patterns to predict how siscowet lake trout visual foraging habitat changes on a daily and seasonal basis. Our model predicted the deepest daytime foraging depths in summer (232.9 m), while the deepest nighttime foraging depths were predicted in winter (32.1 m). Collectively, the findings of these visual studies allow for the improvement of foraging models as well as defining foraging habitat that describes when and where fish may forage.

University Digital Conservancy

Browse by Subject

Browsing by Subject "Vision"