Imagine you draw a typical bedroom, your choice of objects is likely to depend on visual occurrence statistics (i.e. the objects present in previously encountered bedrooms) and semantic relations between objects and scenes (i.e. the semantic relationship between the bedroom and its constituent objects). To investigate how these two factors contribute to the composition of typical scene drawings, we analysed 1192 drawings of six indoor scene categories, obtained from 303 participants. For each object featured in the drawings, we estimated its visual occurrence frequency from the ADE20K dataset, and its semantic relatedness to the scene concept from a word2vec model. Across all scenes of a given category, generalized linear models revealed that visual and conceptual factors both predicted the likelihood of an object featuring in the scene drawings, with a combined model outperforming both single-factor models. We further computed the visual and semantic specificity of objects for a given scene, that is, how diagnostic an object is for the scene. Object specificity offered only weak predictive power when predicting the selection of objects, yet even infrequently drawn objects remained diagnostic of their scenes. Taken together, we show that visual and conceptual factors jointly shape the composition of typical scene drawings.
The most important objects - from predators to projectiles - are often those that are moving. Accordingly, visual systems are specialized for object tracking, and many models of such processing involve continually re-identifying surface features across space and time (as you might follow a tiger by keeping track of its orange stripes). Here, in contrast, we show across four experiments - and in phenomenologically compelling demonstrations - that people also spontaneously perceive and track change-defined objects, with no enduring surface properties from moment to moment. Observers viewed regular grids filled with hundreds of small elements (e.g. randomly oriented crosses). Change-defined objects were implemented by having an element change (e.g., from one random orientation to another random orientation), with these changes propagating through space and time. Observers still detected the 'second-order' motion in such displays - despite the lack of persisting features, and without being able to identify the object in any static frame. And beyond motion detection, observers also spontaneously perceived and tracked persisting objects - and were even able to perform multiple-object tracking. We also generalized this phenomenon in several ways, showing that it occurs with other types of changes (e.g., to brightness or shape), and even when different types of changes are haphazardly interleaved (e.g. with a random orientation change to one element followed by a random brightness change to its neighbor, etc.). In essence, these stimuli show how observers can track a tiger that is always fully hidden while moving through tall grass, by tracking variable local changes to the grass itself.
Imitation learning in complex, unstructured environments remains challenging due to the difficulty of grounding perception in physically meaningful representations and the need to model multimodal action distributions. Existing approaches often rely on unstructured pixel-level feature encodings or stochastic latent-variable decoders, which can lead to brittle attention in cluttered scenes. In this work, we present a novel integration of detector-based visual representations with conditional diffusion modeling (DINO + CDP) for real-world robotic imitation learning. Our framework utilizes a DINO object detection transformer to extract spatially-grounded object-query embeddings that serve as the conditioning signal for a diffusion-based policy. A primary contribution of this work is the systematic quantification of how scene complexity-measured via image entropy-affects robotic policy performance. By comparing rigid-object baselines with complex biological plant scenes, we demonstrate that organic morphology induces a measurable increase in pixel-level uncertainty that degrades standard pixel-centric models. Our results show that DINO + CDP mitigates this degradation by grounding action generation in stable object-level features. We evaluate our approach using a fully real-world manipulation dataset collected without simulation or synthetic pre-training. To isolate the impact of our architectural choices, we conduct a comparative study within a unified framework against convolutional (CNN-MLP), transformer-patch (ViT), and latent-variable (DETR + CVAE) variants. Experimental results in a robotic-arm biocell setup demonstrate that object-query-conditioned diffusion significantly improves task success rates, produces smoother trajectories, and exhibits superior robustness to high-entropy visual inputs, establishing a scalable pathway for imitation learning in challenging agricultural domains.
Orchestration of infrastructure and data resources for research on sensitive data in Trusted Research Environments (TREs) is complex and time-consuming, with fragmented processes for data governance, identity management, disclosure checks, and resource provisioning. Research Object Crates (RO-Crates) aim to make metadata and analytical outputs FAIR (findable, accessible, interoperable, reusable), but current approaches neither record infrastructure requirements in RO-Crates nor use them for automated provisioning. CR8TOR addresses this challenge by extending standardised metadata models (Five-Safes RO-Crate) to take an active role in creating project-dependent infrastructure. Built on the Kubernetes Operator pattern within K8TRE (a Kubernetes-native, cloud-agnostic TRE implementation), CR8TOR demonstrates how contextual project information modelled in RO-Crate-like schemas can provide machine-readable representations of research resource requirements. By extending the Kubernetes API with custom resource definitions (CRDs), these resource descriptions sit alongside traditional research object metadata while maintaining FAIR principles, acting as computable units that can be recreated when required. CR8TOR supports automated data ingress from external sources, project governance controls, data provenance tracking, and on-demand reconciliation of project resources including analytics workspaces, user accounts, and access/network policies. Deployed across cloud and on-premises infrastructure supporting multi-institution collaborations, the same declarative models can be shared across deployments, enabling reproducible and portable research settings. This work demonstrates how provisionable metadata may strengthen methodological foundations of federated research by allowing TRE infrastructure specifications-including workspace configuration, controlled access rules, and project-specific identities-to be described, shared, and reproduced with clarity, consistency, and automation, ultimately lowering operational costs while improving governance and reproducibility.
A key challenge in visual object recognition is developing models that generalize from limited data while maintaining transparency in their decision making. We propose a biologically inspired model that addresses both issues by classifying images based on transformation-invariant local shape key features. Following the principles of the brain's what and where pathways, each feature is encoded by an image patch and its relative location in polar coordinates, enabling interpretable and robust comparisons between inputs and class prototypes. To mimic human concept learning, prototypes are selected using clustering, improving representativeness and generalization. Results show that our model achieves human-comparable performance, with an error rate between 1% and 2% on the MNIST data set when all training images are used as prototypes. In data-limited scenarios, where only a small number of prototypes are selected, our model consistently outperforms convolutional neural networks (CNNs). To evaluate out-of-distribution generalization, we use prototypes from MNIST and test both models on the ETL-1 data set, which differs in data distribution. Although CNN accuracy drops significantly under these conditions, our model maintains high accuracy, even with few prototypes, demonstrating strong robustness and greater capacity to generalize to unseen distributions, bringing it closer to human-like recognition capabilities.
Young children struggle in situations involving mere possibility, such as when a target object could be in one of two possible locations. This difficulty is often interpreted as a general limitation on children's ability to reason about possibility. We tested the hypothesis that a source of children's difficulty may be their early-emerging object representational architecture, which necessarily represents where an object is, but only optionally represents what an object is. Two- to 4-year-old US children (n = 138) were tasked with finding a target object in one of several possible containers, and we manipulated whether the location or the identity of the target was uncertain. At all ages, children were better able to find the target when there was uncertainty about "what" versus uncertainty about "where". We suggest that children's object representational architecture can support uncertainty about identity but not location, and discuss implications for the development of possibility reasoning.
People learn perception-action coupling through repeated interactions with environments. With extensive practice, perception-action coupling becomes automatic, and instructions that contradict such coupling are often ignored, suggesting that the coupling becomes obligatory. One theoretical question is how perception-action coupling in learning contexts generalizes to other contexts. Spatial updating, the process by which people revise their spatial relationship to objects during locomotion, is considered both automatic and obligatory in physical environments, making it ideal for studying coupling generalization. This study examined whether spatial updating remains obligatory in immersive virtual and mixed-reality environments. Participants learned an object array from one orientation and performed judgments of relative direction from two imagined and two actual headings, same as or opposite to the learning heading. Spatial updating was indexed by superior pointing performance when imagined and actual headings were aligned rather than when they were misaligned, an aligned-imagined effect. In Experiments 1 and 2, participants learned virtual objects in immersive virtual environments, with the virtual room either removed or retained during testing. In Experiments 3 and 4, participants learned a mixed layout of real and virtual objects. In Experiment 5, participants learned real objects in a physical environment. Despite instructions to suppress spatial updating, sensorimotor alignment effects appeared in Experiments 3, 4, and 5, but not in Experiments 1 and 2. These findings suggest that spatial updating is not obligatory in immersive virtual environments but remains obligatory in mixed-reality and physical environments. More broadly, obligatory coupling may become less obligatory in transfer contexts, depending on context similarity. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
Conservation conditions assessment of cultural heritage objects is a complex task, requiring the distinction between original and manipulated areas and the identification of materials employed during previous interventions. At the same time, the analysis and detection of exogenous materials should minimise or avoid sampling and prevent sample degradation. In this study, a multi-analytical and multi-scale method is developed to achieve these goals, combining non-invasive hyperspectral imaging in the short-wave infrared region (HSI-SWIR) with micro-invasive FTIR and micro-Raman spectroscopies. The top-down workflow lies on a first macroscopic imaging by HSI-SWIR, allowing to map spatial distributions of both inorganic and organic compounds. However, a deep diagnostic capability with this sole technique is limited. To enhance the spatial distribution estimation of restored portions of the object, a novel ORganic Index (ORI) based on characteristic CH absorptions in the SWIR region is introduced. ORI maps improve the detection of organic materials compared to UV-induced fluorescence imaging and Spectral Angle Mapper classification, enhancing the identification of regions of interest for further analyses. Micro-invasive FTIR and Raman spectroscopies are thus also implemented, enabling molecular-level unambiguous identification of different restoration materials, including polycyanoacrylates, epoxy resins, terpenoid resins, lime-based mortars. Mineralogical specimens embedded in geological matrices are selected as test materials due to their complex three-dimensional geometry and the lack of previous systematic analytical investigations. The proposed methodology represents a reliable non-destructive approach for the study of restoration practices, authenticity assessment, and conservation of complex cultural heritage objects.
Visuospatial deficits are a common and disabling feature of Parkinson's disease (PD) that are linked to dysfunctions of the dorsal visual stream (DVS) and the cholinergic system. However, a body of evidence also points to the possible role of noradrenergic dysfunction, but direct in vivo evidence has remained lacking until the recent development of molecular imaging techniques. Thirty PD patients and thirty matched healthy controls underwent PET imaging using [11C]yohimbine, a new radiotracer now available for human use, to quantify α2-AR availability. Visuospatial cognition was assessed using the MoCA visuospatial subscore, the Visual Object and Space Perception (VOSP) Number Location subtest, and the 15-Objects test. Compared with controls, PD patients showed reduced performance on the VOSP Number Location and 15-Objects tests (both p < 0.05). Voxel-wise analyses revealed reduced [11C]yohimbine binding in DVS regions, including the superior parietal lobule, posterior cingulate cortex, and supramarginal gyri (p_FWE <0.05), but no correlations were observed between α2-AR availability and visuospatial performances. Bayesian analyses provided moderate evidence for the absence of association (BF01 > 3). Despite reduced α2-AR availability within DVS regions, no significant association was observed with visuospatial performance. These findings do not support a simple direct association between DVS α2-AR availability and visuospatial impairment in this cohort of patients with PD.
This study investigated whether haptic contact modulates vestibular-induced self-motion perception and reduces perceptual errors caused by asymmetric vestibular stimulation. Fifteen healthy right-handed participants underwent asymmetric sinusoidal whole-body rotations about the vertical axis in darkness, consisting of half-cycles with equal amplitudes but different velocities, a paradigm known to induce errors in the localization of a previously memorized visual target. This error, quantified as the final position error (FPE), mainly results from reduced self-motion perception during the slower half-cycle. Under vestibular-only conditions, a large FPE was observed after four cycles. To investigate the influence of haptic input, participants touched a spatially fixed object with either the right or the left hand during asymmetric rotation The object was placed either at the body midline or in the right or left hemispace. Haptic contact significantly reduced the vestibular induced FPE, by enhancing perception of the slow self-motion, likely through proprioceptive input from the upper limb. The strongest improvement was observed when participants used the right hand to contact an object in the right hemispace, with smaller reductions occurring with right-hand contact at the midline or in the left hemispace. The effect was less pronounced with left-hand contact. Importantly, although haptic contact markedly reduced vestibular-induced perceptual errors, it did not abolish the underlying vestibular adaptive process responsible for the FPE. In conclusion, haptic contact significantly mitigates vestibular-induced self-motion misperception, with effects depending on both hand laterality and contact location.
Traumatic brain injury (TBI) can result in long-lasting cognitive impairment. Accumulating clinical evidence indicates that outcomes often differ according to biological sex. Despite this, preclinical studies remain predominantly male-focused, which limits our understanding of sex-specific vulnerabilities. In this study, we investigated long-term memory outcomes in male and female rats following moderate-to-severe fluid percussion injury, with an additional examination of estrous cycle phases in females. Behavioral performance was assessed using the open field test, novel object location (NOL), novel object recognition (NOR), and the Barnes maze (BM) test. We found that, independent of injury, females exhibited greater locomotor activity than males. TBI significantly impaired performance in NOL and NOR tasks, with sex-dependent patterns of deficit. Female rats exposed to TBI showed greater impairment in novelty-driven memory tasks, including significantly poorer object recognition, than TBI-exposed males. In contrast, BM testing revealed more subtle effects, with TBI causing both sexes to shift towards nonspatial search strategies during the probe trial and males presenting greater disruption in learning-dependent spatial navigation. The phase of the estrous cycle during behavioral testing did not significantly influence the results. Overall, our findings suggest that sex differences in cognitive outcomes following TBI appear to be domain-specific. These results emphasize the importance of incorporating sex as a biological variable in preclinical TBI models, which could help to explain the variability in cognitive outcomes across studies.
Comparisons are fundamental to science: experiment against model, one organism against another, a system against itself across time. Because many systems, from brains to climate, are characterized by how they evolve in time, it is a natural goal to compare their dynamics. Dynamical systems comparison is well defined, but has been intractable for nonlinear, high-dimensional, noisy, and partially observed data. As a result, standard comparison metrics have focused on the geometry or topology of data. Here we present Dynamical Similarity Analysis (DSA), a class of methods to compare systems by their temporal evolution. Its foundation is Koopman Operator theory, which recasts nonlinear systems as linear operators. We estimate these operators from data, then compare the operators across systems. The computation is fast, scalable, and robust to noise and partial observation. It is also differentiable. DSA identifies dynamical structure that geometric and topological methods miss. It matches recordings from the head direction circuit to ring attractor models. It shows that macaque motor cortex dynamics for two reaching tasks drift apart across years despite preserved behavior, and that primary motor cortex breaks from premotor cortex as movement begins. As an optimization objective, it induces neural networks to learn never-before hypothesized solutions that run counter to their inductive biases. Thus, DSA transforms the dynamics of a system into an object that can be measured, compared, and optimized.
This study reports on an empirical study, focusing on the declarative and procedural knowledge in English-speaking learners' acquisition of the verbal identity condition in their second language Chinese grammars. While subjects and objects in English sentences are generally overt, both subjects and objects in Chinese can be 'covert'. It is argued in this article that the absence of subjects and objects in affirmative answers to Chinese yes-no questions arises from TP ellipsis under a verbal identity condition (Simpson, 2014). A cross-modal self-paced reading task (SRT) and an acceptability judgment task (AJT) are involved in the study, and the results of the AJT show that all learner groups, like native Chinese speakers, judge answers with covert arguments under the verbal identity condition as more natural than those not under such a condition. In the SRT, however, beginner learners, but not intermediate and advanced learners, fail to show sensitivity to the verbal identity condition. These findings align with the Declarative/Procedural Model (Ullman, 2005, 2015), indicating that beginners rely on declarative, rule-based knowledge to evaluate grammaticality but lack proceduralized capacity to process covert arguments in real time. With increased exposure and proficiency, learners proceduralize grammatical computations, such as lexical identity checking, movement, and ellipsis, which enable them to acquire native-like sensitivity to the verbal identity condition in their procedural as well as declarative knowledge.
BackgroundIt has been demonstrated that grating interferometers can be utilized to enhance the X-ray imaging contrast of low-density and weakly absorbing objects.ObjectiveIn this study, the design, instrumentation and validation of a grating based laboratory 5.41 keV X-ray phase-contrast microscope with zone-plate are reported.MethodsThis dedicated microscope was carefully designed by coupling the zone-plate imaging method and the grating interferometer imaging method. Specially fabricated X-ray optics were integrated along with a certain X-ray tube source illumination module, motion stages and X-ray detection module. The system's spatial resolution was characterized with the modulation transfer function (MTF) analysis and the power spectral density (PSD) analysis. The absorption-contrast and phase-contrast images were extracted from the acquired phase stepping dataset. The splitting and superposition behavior of the phase-contrast signal was evaluated by comparing the experimental results with the numerical simulation. CT projections of a mitochondria-targeted, antibody-stained Jurkat T cell were pre-processed by geometric registration before performing the standard filtered back-projection (FBP) reconstruction.ResultsThe absorption imaging results of the Siemens star phantom demonstrated that this microscope achieved a spatial resolution of 35.0 nm based on MTF analysis, while the minimum detectable structure size was 28.5 nm based on PSD analysis. The phase-contrast imaging of a 5.00μm diameter polystyrene microsphere demonstrated the splitting phenomenon (8.54μm on the sample plane) of the ±1st order components, and both the splitting distance and line profiles agreed well with the theoretical imaging model. In the reconstructed phase-contrast CT images, the mitochondria could be clearly identified.ConclusionsThis study reported the development of a grating based laboratory 5.41 keV X-ray phase-contrast microscope with zone-plate. Results showed that it could significantly enhance the imaging performance of low-density objects such as polystyrene microsphere and biological cells.
We show that oscillating (real-scalar) boson stars generically host an oscillating radial caustic. Sources near this caustic cross it every half period, thereby producing periodic caustic-crossing lensing. The resulting observables are phase locked to the lens oscillation: image-pair creation or annihilation, changing image morphology, achromatic photometric spikes, and astrometric motion. This signal provides a distinctive target for time-domain astronomy, and its detection would reveal an intrinsically time-dependent compact dark-sector object. Event-number estimates indicate a measurable discovery space with current astrometric and high-cadence photometric surveys, while null searches would constrain the abundance of such objects as dark matter. The predictions rely only on the dynamics of real-scalar condensates and extend naturally to self-interacting real scalars, including axionlike particles, and to ultralight vector bosons.
Emotion regulation and memory control are indispensable for well-being, yet their interplay remains poorly understood. In this preregistered electroencephalogram (EEG) study, we investigated how emotion regulation influences the subsequent control of negative memories. Participants learned object-aversive scene pairings and were then instructed to either naturally watch or actively reappraise the aversive scenes. They subsequently completed an EEG-based Think/No-Think (TNT) task, where they either retrieved or suppressed the retrieval of aversive scenes in response to Think or No-Think object cues while reporting any intrusions they experienced. Different from our initial hypothesis, preregistered analyses showed that during retrieval suppression, reappraised memories were associated with larger centro-parietal P300 than watched scenes. Consistently, in the post-TNT recall test, participants recalled more details and reported lower negative emotions related to these reappraised scenes than the watched scenes. These results suggest that cognitive reappraisal may preserve the episodic content of negative memories while lessening their emotional impact during recall. This finding holds significant clinical potential, as it could enable individuals to retain valuable information from adverse experiences without being overwhelmed by distress. Additionally, it may help address reduced memory specificity in affective disorders, offering a promising avenue for therapeutic intervention.
Referring remote sensing image segmentation (RRSIS) enables the precise delineation of regions within remote sensing imagery through natural language descriptions, serving critical applications in disaster response, urban development, and environmental monitoring. Despite recent advances, current approaches face significant challenges in processing aerial imagery due to complex object characteristics including scale variations, diverse orientations, and semantic ambiguities inherent to the overhead perspective. To address these limitations, we propose DiffRIS, a novel framework that harnesses the semantic understanding capabilities of pre-trained text-to-image diffusion models for enhanced cross-modal alignment in RRSIS tasks. Our framework introduces two key innovations: a context perception adapter (CP-adapter) that dynamically refines linguistic features through global context modeling and object-aware reasoning, and a progressive cross-modal reasoning decoder (PCMRD) that iteratively aligns textual descriptions with visual regions for precise segmentation. The CP-adapter bridges the domain gap between general vision-language understanding and remote sensing applications, while PCMRD enables fine-grained semantic alignment through multi-scale feature interaction. Comprehensive experiments on three benchmark datasets-RRSIS-D, RefSegRS, and RISBench-demonstrate that DiffRIS consistently outperforms existing methods across all standard metrics, establishing a new state-of-the-art for RRSIS tasks. The significant performance improvements validate the effectiveness of leveraging pre-trained diffusion models for remote sensing applications through our proposed adaptive framework.
Epilepsy is a neurological disorder characterized by excessive neuronal firing, frequently originating in the hippocampus. Hyperpolarization-activated cyclic nucleotide-gated channel-1 (HCN1) regulates neuronal excitability and resting membrane potential, yet its role in seizure progression remains unclear. Cannabidiol (CBD), an effective anticonvulsant, may exert part of its effects through HCN1. This study investigated the contribution of HCN1 to seizure progression, synaptic plasticity, and CBD-mediated neuroprotection. Rats were implanted with stimulation electrodes in the perforant path (PP) and recording electrodes with a guide cannula in the dentate gyrus (DG). One week later, lentiviral shRNA-HCN1 was injected into the DG, followed by PP electrical kindling. CBD (100 ng/2 μL) was administered every other day in shRNA-HCN1-treated or non-manipulated animals. Seizure severity was assessed using Racine's scale. Synaptic transmission, paired-pulse plasticity, and long-term potentiation (LTP) were evaluated by extracellular field recordings, HCN1 function by whole-cell patch-clamp recordings of Ih (Hyperpolarization-activated current), HCN1 expression by RT-qPCR, and recognition memory using the novel object recognition (NOR) test. Kindling reduced HCN1 mRNA expression, which was further decreased by shRNA-HCN1. HCN1 knockdown accelerated seizure progression, prolonged after-discharge duration, increased spike activity, reduced the sag ratio, and impaired synaptic transmission, paired-pulse plasticity, LTP, and object recognition memory in fully kindled rats. CBD significantly attenuated these electrophysiological and recognition memory deficits, although its protective effects were partially reduced following HCN1 knockdown. These findings indicate that HCN1 contributes to seizure progression and hippocampal dysfunction, while CBD exerts anticonvulsant and neuroprotective effects through both HCN1-dependent and HCN1-independent mechanisms.
Fused filament fabrication 3D printers have become popular for at-home use. While these printers enable the convenient creation of 3D objects, their potential inhalation hazards are not widely known among the public. 3D printer feedstocks, known as filaments, consist of a variety of plastics, including acrylonitrile butadiene styrene (ABS), poly-lactic acid (PLA), and polycarbonate (PC), all of which have been linked to the emission of indoor air particles during 3D printing. These particle emissions have also been linked to adverse effects in in-vivo and in-vitro studies. The objective of this study was to characterize particle emissions from seventeen commercially available printing filaments in an enclosed test chamber and explore how particles behaved at different time points during the 3D printing process. Particle number concentrations, particle number emission rates, and size distributions were collected and quantified from a sampling flask using a scanning mobility particle sizer. Particle number concentrations were greatest among ABS filaments, followed by metal-containing PLA, PLA, and PC filaments. Size distributions at different time points showed that particle distributions widened and reached a stable distribution at approximately ten minutes into data collection. On average, particles were also observed to be largest among ABS filaments while emissions from one PLA manufacturer showed consistently larger particle sizes than the other. A relative ranking comparison of average particle number concentrations with results from a recent study [1] that used filaments with the same lot numbers found some similarity, despite different methods for particle capture.
Gastroretentive drug delivery systems aim to prolong gastric residence time, yet their in vivo performance is often inconsistent, largely due to insufficient consideration of mechanically driven gastric emptying. In particular, the role of antral peristalsis in the transport of solid oral dosage forms is underrepresented in current in vitro models. In this study, a next-generation in vitro antrum model was developed to simulate mechanically relevant conditions governing gastric emptying. The system is based on a flexible tubular compartment with mechanically induced peristaltic waves and allows controlled variation of key parameters, including wave velocity (3 mm/s), fluid volume (50-150 mL), inclination (0-40°), and occlusion as residual lumen diameter (1.6-25.6 mm).Using test objects with defined differences in size, density, and deformability, the model demonstrated that transport behavior is primarily governed by the interplay of geometric confinement, deformation, and contact mechanics. Rigid objects were transported once a critical lumen diameter threshold was reached, whereas highly deformable and entangled structures partially resisted peristaltic transport. Observed size-dependent transport behavior was consistent with in vivo findings from relevant literature. The model provides a simple and mechanistically relevant platform for early-stage screening of gastroretentive dosage forms and supports a mechanics-driven understanding of gastroretention.