Visual Domain Research Articles

Geometric Understanding of Discriminability and Transferability for Visual Domain Adaptation.

To overcome the restriction of identical distribution assumption, invariant representation learning for unsupervised domain adaptation (UDA) has made significant advances in computer vision and pattern recognition communities. In UDA scenario, the training and test data belong to different domains while the task model is learned to be invariant. Recently, empirical connections between transferability and discriminability have received increasing attention, which is the key to understand the invariant representations. However, theoretical study of these abilities and in-depth analysis of the learned feature structures are unexplored yet. In this work, we systematically analyze the essentials of transferability and discriminability from the geometric perspective. Our theoretical results provide insights into understanding the co-regularization relation and prove the possibility of learning these abilities. From methodology aspect, the abilities are formulated as geometric properties between domain/cluster subspaces (i.e., orthogonality and equivalence) and characterized as the relation between the norms/ranks of multiple matrices. Two optimization-friendly learning principles are derived, which also ensure some intuitive explanations. Moreover, a feasible range for the co-regularization parameters is deduced to balance the learning of geometric structures. Based on the theoretical results, a geometry-oriented model is proposed for enhancing the transferability and discriminability via nuclear norm optimization. Extensive experiment results validate the effectiveness of the proposed model in empirical applications, and verify that the geometric abilities can be sufficiently learned in the derived feasible range.

TDSCCNet: twin-depthwise separable convolution connect network for change detection with heterogeneous images

The task of change detection (CD) in optical and SAR images is an ever-evolving and demanding subject within the realm of remote sensing (RS). It holds great significance to identify the target areas by using complementary information between the two. Due to the distinct imaging mechanisms employed by optical and SAR sensors, effectively and accurately identifying changing regions can be challenging. To this end, a novel heterogeneous RS images CD network (Twin-Depthwise Separable Convolution Connect Network, TDSCCNet) is proposed in this paper. Image domain transformation is a front-end task, while the back-end employs a single-branch bilayer depthwise separable convolution-connected encoder-decoder to accomplish CD work. Specifically, first, the cycle-consistent adversarial network (CycleGAN) serves to integrate the optical and SAR visual domains, and a consistent feature expression is obtained. Second, the single-branch encoder structure of bilayer depthwise separable convolution is employed to realize change feature extraction. Finally, the multiscale connected decoder reconstructed by change map is utilized to reconstruct the original images and solve the local discontinuities and holes in the binary change map. Multiscale loss is designated for optimizing the global and local effects to alleviate the class imbalance problem. It was tested on four representative datasets from Gloucester, Shuguang Village, Italy and WV-3 datasets with an overall accuracy of 97.36%, 97.01%, 97.62%, and 98.01% respectively. By comparing the existing methods, experimental results confirmed the effectiveness of the proposed method.

Structured Domain Adaptation With Online Relation Regularization for Unsupervised Person Re-ID.

Unsupervised domain adaptation (UDA) aims at adapting the model trained on a labeled source-domain dataset to an unlabeled target-domain dataset. The task of UDA on open-set person reidentification (re-ID) is even more challenging as the identities (classes) do not have overlap between the two domains. One major research direction was based on domain translation, which, however, has fallen out of favor in recent years due to inferior performance compared with pseudo-label-based methods. We argue that domain translation has great potential on exploiting valuable source-domain data but the existing methods did not provide proper regularization on the translation process. Specifically, previous methods only focus on maintaining the identities of the translated images while ignoring the intersample relations during translation. To tackle the challenges, we propose an end-to-end structured domain adaptation framework with an online relation-consistency regularization term. During training, the person feature encoder is optimized to model intersample relations on-the-fly for supervising relation-consistency domain translation, which in turn improves the encoder with informative translated images. The encoder can be further improved with pseudo labels, where the source-to-target translated images with ground-truth identities and target-domain images with pseudo identities are jointly used for training. In the experiments, our proposed framework is shown to achieve state-of-the-art performance on multiple UDA tasks of person re-ID. With the synthetic→real translated images from our structured domain-translation network, we achieved second place in the Visual Domain Adaptation Challenge (VisDA) in 2020.

Visual history and semiotics

This paper investigates the reflection of “visual turn”in semiotic research, the relationship between visual history and semiotics. The aim of the research is to investigate the relationship between visual history and semiotics and to reflect this relationship in scholarly research. The English-language works of such famous researchers and cultural theorists as Ferdinand de Saussure, Michel Foucault, Algirdas Julius Greimas, Jean Francois Lyotard, Jacques Lacan, Ludwig Wittgenstein, Jacques Derrida Jacques Derrida, Umberto Eco, Mi-chael Halliday, Charles Peirce, Noam Chomsky, and others were analyzed in this paper. History does not speak for itself, so the work of аhistorian is always interpretive. The appeal to the visual here is not a turn away from other ways, other meanings of making history. The visual is seen here as a special form of knowledge, requiring its own ways of apprehending it. This engagement with the visual will undoubtedly have a significant impact on our understanding of archives and historiography in general but does not claim to privilege the visual domain as offering a special kind of access to the past

Exploring office comfort and productivity in living labs: A yearlong structural equation modeling study

The relationship between workers' environmental comfort and productivity has been extensively studied over the years, yet real-world investigations remain limited. Living labs provide an opportunity to observe interactions between different comfort domains and their impact on occupants' attitudes and preferences. This study aims to investigate the connection between contextual factors, visual and thermal sensations, and comfort, and how they relate to the self-perceived productivity of office workers in a Living Lab setting, utilizing Structural Equation Modelling (SEM). Comfort and sensation perception surveys were conducted during each session over a year, alongside a sensing system continuously monitoring air temperature, illuminance, and window/door operations in the five offices of a dedicated living lab participating in the study. Subjects working in single offices reported higher productivity than those in shared offices across all comfort conditions. Contextual characteristics, such as the time of day, had a significant impact on defining thermal and visual satisfaction. A significant cross-modal effect was observed between thermal and visual domains, highlighting that evaluating factors in isolation from comfort assessment overlooks critical elements of building physics. These findings underscore the importance of considering the broader context when assessing people's perception of their surroundings and how operational changes in office spaces can enhance employees' productivity judgment.

Noun imageability and the processing of sensory-based information.

The aim of this study was to test whether the availability of internal imagery elicited by words is related to ratings of word imageability. Participants are presented with target words and, after a delay allowing for processing of the word, answer questions regarding the size or weight of the word referents. Target words differ with respect to imageability. Results show faster responses to questions for high imageability words than for low imageability words. The type of question (size/weight) modulates reaction times suggesting a dominance of the visual domain over the physical-experience domain in concept representation. Results hold across two different languages (Czech/German). These findings provide further insights into the representations underlying word meaning and the role of word imageability in language acquisition and processing.

Event-related potential (ERP) evidence for visual processing differences in children and adults with cystinosis (CTNS gene mutations)

BackgroundCystinosis, a rare lysosomal storage disease caused by mutations in the CTNS gene, is characterized by cystine crystallization and accumulation within multiple tissues, including kidney and brain. Its impact on neural function appears mild relative to its effects on other organs during early disease, but since therapeutic advances have led to substantially increased life expectancy, neurological implications are of increasing interest, necessitating deeper understanding of the impact of cystinosis on neurocognitive function. Behavioral difficulties have been reported in cystinosis in the visual domain. Very little is known, however, about how the brains of people living with cystinosis process visual information. This is especially interesting given that cystine accumulation in the cornea and posterior ocular structures is a hallmark of cystinosis.MethodsHere, high-density scalp electrophysiology was recorded to visual stimuli (during a Go/No-Go task) to investigate visual processing in individuals with cystinosis, compared to age-matched controls. Analyses focused on early stages of cortical visual processing.ResultsThe groups differed in their initial cortical response, with individuals with cystinosis exhibiting a significantly larger visual evoked potential (VEP) in the 130–150 ms time window. The groups also differed in the associations between neural responses and verbal abilities: While controls with higher IQ scores presented larger neural responses, that relationship was not observed in cystinosis.ConclusionsThe enlarged VEP in cystinosis could be the result of cortical hyperexcitability and/or differences in attentional engagement and explain, at least partially, the visual and visual-spatial difficulties described in this population.

Reduced tactile sensitivity is associated with mild cognitive impairment

Sensory impairment has been related to age-associated cognitive decline. While these associations were investigated primarily in the auditory and visual domain, other senses such as touch have rarely been studied. Thus, it remains open whether these results are specific for particular sensory domains, or rather point to a fundamental role of sensory deficits in cognitive decline. Data from 31 participants with mild cognitive impairment (MCI), 46 participants with frailty, and 23 non-clinical control participants (NCCs) were included. We assessed sensory function using visual acuity and contrast sensitivity, hearing threshold, and mechanical detection threshold. Cognitive function in participants with MCI was assessed using associative memory performance. Group differences on sensory thresholds were tested using analyses of covariance with age, sex, and years of education as covariates. Associations between measures within participants with MCI were evaluated using Spearman correlations. We found a significant difference in mechanical detection threshold between the groups (p<0.001, η2=0.18). Participants with MCI showed significantly reduced tactile sensitivity compared to participants with frailty and NCCs. In participants with MCI, lower associative memory performance was significantly related to reduced tactile sensitivity (rs=0.39, p=0.031) and auditory acuity (rs=0.41, p=0.022). Our results indicate that reduced tactile sensitivity is related to cognitive decline. Prospective studies should investigate the age-related alterations of multimodal sensory processes and their contribution to dementia-related processes. Deutsche Forschungsgemeinschaft (FL 156/41-1) and a grant of the Hector-Stiftung II, Weinheim, Germany.

Multisensory Hallucinatory Experiences in Migraine: A Preliminary Basis for Olfactory, Somatic-Tactile, and Gustatory Auras.

Visual auras in migraine have been extensively studied, but less is known about unusual experiences in other sensory domains, including whether they should be diagnostically considered as part of aura symptoms. This study aimed to examine the prevalence of multisensory aura experiences in migraine and their phenomenologic and clinical correlates. Respondents were 729 participants with probable migraine, who completed an online study examining unusual sensory experiences. These comprised aura experiences in the auditory, visual, olfactory, somatic-tactile, and gustatory domains. Basic demographic and clinical information and migraine symptomatology were also collected. To facilitate groupwise comparisons, participants with probable migraine were divided into those with and without (visual) aura experiences. Endorsement of visual aura experiences was the highest (42.1%), whether in a single modality (44.2%) or multiple (55.8%) modalities, followed by somatic-tactile (32.0%), gustatory (21.9%), olfactory (18.6%), and auditory (11.0%) domains. Phenomenologic similarities, for instance, in frequency, personification, and controllability, existed across sensory domains. Somatic-tactile and gustatory auras conversely exhibited greater duration and negative emotional valence. Participants with probable migraine with visual aura tended to report significantly more severe migraine symptoms relative to those with nonvisual or no aura. Our findings provide preliminary indication that unusual olfactory, somatic-tactile, and gustatory experiences in migraine are common and could be clinically significant as aura symptoms. Increased clinician and patient awareness and effective management of these symptoms are essential for a holistic therapeutic approach to migraine.

Biomarker (Metabolic and Lipids) Determinants of Dementia in Older Adult Population from Nigeria: A Cross Sectional Analysis

AbstractBackgroundDementia is determined by underlying biological processes. There is a paucity of studies from Nigeria concerning biomarkers in dementia. We present baseline data from longitudinal study on association of lipid and metabolic biomarkers on dementia.MethodThree hundred older adults (65‐89 years; mean age: 70.30; female: 202) without stroke participated were assessed using adapted Unified Data Set (UDS 3) neuropsychological test battery. Diagnoses was made using the Jack and Bondi (1961) method (normal cognition, borderline or severe impairment/dementia). Blood samples were used to analyze lipid (total cholesterol, triglycerides, HDL, LDL, and VLDL) and sugar. Hypertension, pulse and BMI were obtained in this cross sectional study. Partial correlations controlling for age, sex, education, and marital status were used to test for associations between biomarkers and cognitive status across entire sample. MANOVA was used to analyze differences in biomarker categories on UDS3 performance.ResultThere were no significant associations between dementia status and metabolic biomarkers (systolic BP, diastolic BP, diabetes, BMI and pulse), though diabetes and BMI showed small effect sizes (0.10).No significant association between lipids biomarkers (total cholesterol (TC), triglycerides (TG), HDL, LDL, VLDL) and dementia were found. Also, no significant associations were found between lipids and metabolic biomarkers on cognitive domains, except for HDL on visual spatial domain (‐0.15).There were small effect sizes (0.10 to 0.12) for TC and HDL on attention and TG and VLDL on language. There were small effect sizes (between 0.15 to 0.18) for BMI on attention; systolic BP, diabetes and BMI on visual spatial; systolic BP, diabetes, BMI and pulse on processing speed, as well as pulse on language. No significant differences were seen on cognitive measures for those diagnosed with or without diabetes or hypertention. When comparing high/low HDL, significant differences were found only for language (0.57).ConclusionOur findings stand in contrast to prevailing western studies on the roles of metabolic syndrome and lipids on cognitive status. Though there are pockets of small effect sizes, there was little evidence for metabolic and lipid profiles as determinants of dementia and cognitive performance in older Nigerian adults.

Neural mechanisms of sequential dependence in time perception: the impact of prior task and memory processing.

Our perception and decision-making are susceptible to prior context. Such sequential dependence has been extensively studied in the visual domain, but less is known about its impact on time perception. Moreover, there are ongoing debates about whether these sequential biases occur at the perceptual stage or during subsequent post-perceptual processing. Using functional magnetic resonance imaging, we investigated neural mechanisms underlying temporal sequential dependence and the role of action in time judgments across trials. Participants performed a timing task where they had to remember the duration of green coherent motion and were cued to either actively reproduce its duration or simply view it passively. We found that sequential biases in time perception were only evident when the preceding task involved active duration reproduction. Merely encoding a prior duration without reproduction failed to induce such biases. Neurally, we observed activation in networks associated with timing, such as striato-thalamo-cortical circuits, and performance monitoring networks, particularly when a "Response" trial was anticipated. Importantly, the hippocampus showed sensitivity to these sequential biases, and its activation negatively correlated with the individual's sequential bias following active reproduction trials. These findings highlight the significant role of memory networks in shaping time-related sequential biases at the post-perceptual stages.

More Reliable Neighborhood Contrastive Learning for Novel Class Discovery in Sensor-Based Human Activity Recognition

Human Activity Recognition (HAR) systems have made significant progress in recognizing and classifying human activities using sensor data from a variety of sensors. Nevertheless, they have struggled to automatically discover novel activity classes within massive amounts of unlabeled sensor data without external supervision. This restricts their ability to classify new activities of unlabeled sensor data in real-world deployments where fully supervised settings are not applicable. To address this limitation, this paper presents the Novel Class Discovery (NCD) problem, which aims to classify new class activities of unlabeled sensor data by fully utilizing existing activities of labeled data. To address this problem, we propose a new end-to-end framework called More Reliable Neighborhood Contrastive Learning (MRNCL), which is a variant of the Neighborhood Contrastive Learning (NCL) framework commonly used in visual domain. Compared to NCL, our proposed MRNCL framework is more lightweight and introduces an effective similarity measure that can find more reliable k-nearest neighbors of an unlabeled query sample in the embedding space. These neighbors contribute to contrastive learning to facilitate the model. Extensive experiments on three public sensor datasets demonstrate that the proposed model outperforms existing methods in the NCD task in sensor-based HAR, as indicated by the fact that our model performs better in clustering performance of new activity class instances.

Universal Representations: A Unified Look at Multiple Task and Domain Learning

AbstractWe propose a unified look at jointly learning multiple vision tasks and visual domains through universal representations, a single deep neural network. Learning multiple problems simultaneously involves minimizing a weighted sum of multiple loss functions with different magnitudes and characteristics and thus results in unbalanced state of one loss dominating the optimization and poor results compared to learning a separate model for each problem. To this end, we propose distilling knowledge of multiple task/domain-specific networks into a single deep neural network after aligning its representations with the task/domain-specific ones through small capacity adapters. We rigorously show that universal representations achieve state-of-the-art performances in learning of multiple dense prediction problems in NYU-v2 and Cityscapes, multiple image classification problems from diverse domains in Visual Decathlon Dataset and cross-domain few-shot learning in MetaDataset. Finally we also conduct multiple analysis through ablation and qualitative studies.

Insulation aging condition assessment of transformer in the visual domain based on SE-CNN

The aging of transformer oil–paper insulation (OPI) contributes significantly to transformer failure and reduced service life. Therefore, reliable and precise evaluation of the aging status of transformer insulation condition is crucial in averting transformer failure. This article proposes a novel deep learning-based aging condition assessment technique of transformers by combining FDS (Frequency domain spectroscopy) and squeeze-and-excitation-enabled convolutional neural network (SE-CNN). An intrinsic product-based transformation using a Gramian angular field is employed to dimensionally expand the 1-D FDS signals acquired from OPI samples into 2D images. Each feature extraction channel in the SE-CNN-based aging diagnosis framework is assigned a specific weight, enabling the model to concentrate on the most relevant features while discarding the rest. Furthermore, the effect of moisture variation on the model is also considered under three different conditions to ensure the consequence of compensation for the effect of moisture on FDS. A comparison with traditional machine learning models has also been made to ensure the viability of the proposed methodology. Additionally, the model is interpreted by creating explanation weight plots of the feature coefficients using a local interpretable model-agnostic explanations (LIME) method in Python. The findings demonstrate that the novel aging diagnosis framework visually assesses the transformer’s OPI system with the highest accuracy of 99.12% in the test set. Furthermore, it facilitates the automatic extraction of deep aging features, thus providing better generalization and convergence capability.

Temporal hierarchy of observed goal-directed actions

During social interactions, we continuously integrate current and previous information over varying timescales to infer other people's action intentions. Motor cognition theories argue for a hierarchical organization of goal-directed actions based on temporal scales. Accordingly, transient motor primitives are represented at lower levels of the hierarchy, a combination of primitives building motor sequences at subordinate levels, and more stable overarching action goals at superordinate levels. A neural topography of hierarchal timescales for information accumulation was previously shown in the visual and auditory domains. However, whether such a temporal hierarchy can also account for observed goal-directed action representations in motor pathways remains to be determined. Thus, the current study examined the neural architecture underlying the processing of observed goal-directed actions using inter-subject correlation (ISC) of fMRI activity. Observers (n = 24) viewed sequential hand movements presented in their intact order or piecewise scrambled at three timescales pertaining to goal-directed action evolution (Primitives: ± 1.5 s, Sub-Goals: ± 4 s, and High-Goals: ± 10 s). The results revealed differential intrinsic temporal capacities for integrating goal-directed action information across brain areas engaged in action observation. Longer timescales (> ± 10 s) were found in the posterior parietal and dorsal premotor compared to the ventral premotor (± 4 s) and anterior parietal (± 1.5 s) cortex. Moreover, our results revealed a hemispheric bias with more extended timescales in the right MT+, primary somatosensory, and early visual cortices compared to their homotopic regions in the left hemisphere. Our findings corroborate a hierarchical neural mapping of observed actions based on temporal scales of goals and provide further support for a ubiquitous time-dependent neural organization of information processing across multiple modalities.

Examining safe spaces for pedestrians and e-bicyclists at urban crosswalks: An analysis based on drone-captured video

Despite numerous theoretical and empirical studies exploring the spatial needs of road users, a significant gap remains in validating these findings with extensive real-world data sets. This study presents the idea of “safe spaces,” outlining the buffer zones that both walkers and e-bicyclists maintain when crossing streets, while also taking safety and psychological well-being into consideration. We used drones to gather the study's trajectory data at three significant crossings in Xi'an, China. Multi-coordinate system transformation enabled us to determine the relative positions of individuals and moving objects within their visual domain. Relative position frequency heat maps were generated to explore safe distance demands in different directions. The safety space was then fitted using the least squares method. Our research demonstrates that whereas e-bicyclists maintain semi-elliptical safe spaces at street crossings, walkers maintain semi-circular safe spaces, and the sizes of these zones increase in direct proportion to their relative speeds. However, the safe space size oscillates within a defined range above a particular speed threshold. Notably, e-bicyclists require larger safety distances forward and are more sensitive to speed variations. Our knowledge of the dynamics of safe spaces for walkers and e-bicyclists at street crossings is improved by this work, with emphasis on the role of speed and pedestrian numbers in shaping these spaces. The incorporation of real-world data from drone footage significantly strengthens the validity and reliability of our findings, bridging a crucial empirical gap in the existing literature. Additionally, this paper introduces a crowding coefficient based on safe space and proposes a new method using safe space to aid traffic conflict metrics PET, providing valuable insights for identifying crowd congestion levels and capturing traffic conflict events. The practical implications of our findings extend to urban planning, traffic management, and safety of vulnerable road users. Ultimately, this research contributes to the development of safer and more efficient urban environments for all road users.

Style-Hallucinated Dual Consistency Learning: A Unified Framework for Visual Domain Generalization

Style-Hallucinated Dual Consistency Learning: A Unified Framework for Visual Domain Generalization

Analisis Visual Personal Branding Sisca Kohl di Instragram

The growth of creator content is a sign of how social media is developing, especially Instagram, which ranks fourth as social media with the most users. The presence of Instagram as a new medium is also able to build one's personal branding. Therefore, content creators must try hard to establish their brand image to be known. This research is a descriptive qualitative research. The descriptive method is used to find meaning, understanding, understanding of a phenomenon or event by being directly or indirectly involved in the object under study contextually and thoroughly. In this study, we will examine how the visualization of Sisca Kohl's personal branding on her Instagram account @siscakohl. This research refers to three visual research domains production site, self site, and audience site using Gillian Rose's visual methodology. The results of the study show that based on the results of analysis through content uploaded by Sisca on the @siscakohl Instagram account, it fulfills 5 personal branding laws out of 8 personal branding laws according to Montoya. namely specialization (the law of specialization), leadership (the law of leadership), difference (the law of distinctiveness), the law of visibility, unity (the law of unity). Of the 5 personal branding laws from Montoya, the most prominent are specialization (The Law of Specialization) and difference (the law of distinctiveness).

Neural correlates of cognitive control deficits in pediatric mild traumatic brain injury.

There is a growing body of research showing that cerebral pathophysiological processes triggered by pediatric mild traumatic brain injury (pmTBI) may extend beyond the usual clinical recovery timeline. It is paramount to further unravel these processes, because the possible long-term cognitive effects resulting from ongoing secondary injury in the developing brain are not known. In the current fMRI study, neural processes related to cognitive control were studied in 181 patients with pmTBI at sub-acute (SA; ~1 week) and early chronic (EC; ~4 months) stages post-injury. Additionally, a group of 162 age- and sex-matched healthy controls (HC) were recruited at equivalent time points. Proactive (post-cue) and reactive (post-probe) cognitive control were examined using a multimodal attention fMRI paradigm for either congruent or incongruent stimuli. To study brain network function, the triple-network model was used, consisting of the executive and salience networks (collectively known as the cognitive control network), and the default mode network. Additionally, whole-brain voxel-wise analyses were performed. Decreased deactivation was found within the default mode network at the EC stage following pmTBI during both proactive and reactive control. Voxel-wise analyses revealed sub-acute hypoactivation of a frontal area of the cognitive control network (left pre-supplementary motor area) during proactive control, with a reversed effect at the EC stage after pmTBI. Similar effects were observed in areas outside of the triple-network during reactive control. Group differences in activation during proactive control were limited to the visual domain, whereas for reactive control findings were more pronounced during the attendance of auditory stimuli. No significant correlations were present between task-related activations and (persistent) post-concussive symptoms. In aggregate, current results show alterations in neural functioning during cognitive control in pmTBI up to 4 months post-injury, regardless of clinical recovery. We propose that subacute decreases in activity reflect a general state of hypo-excitability due to the injury, while early chronic hyperactivation represents a compensatory mechanism to prevent default mode interference and to retain cognitive control.

Using an ‘Ethnogram’ to visualize talk in the classroom

ABSTRACT Ethnograms are used to visualize data gathered from students’ and teachers’ talk in the classrooms of a secondary school in the UK. A sample of talk data is portrayed in three example ethnograms. The ethnograms transform non-visual data into the visual domain for the purpose of analysis using visualization and abstraction. Using ethnograms the sound of talk, as it had occurred during lessons, became seen and thus visualized which students talked most or made the most noise, where teachers positioned themselves and the relationships between students’ talk and teachers’ positions in classrooms. Through visualizing talk, it became possible to see space in the classroom which was louder or quieter and identify students who were silent. Visualizing talk led to alternate perspectives and interpretations of the data and surprising findings to be surfaced. Ethnograms are therefore posited as a potential method for researchers interested in portraying data for further post-collection analysis or to see, for example, sensory data such as mood, emotion or smell. Ethnograms are shown to be an accessible and viable qualitative research method particularly useful for researchers who wish to qualitatively visualize the social interactions and behaviours of people for interpretation.