Modern multi-view clustering (MVC) is dominated by two paradigms: multi-view fusion and pseudo-label-guided learning. Pseudo-labeling methods can suffer from confirmation bias; their reliance on a fixed-granularity supervision from an initial clustering can cause learned embeddings to drift from the data's true structure and lose discriminative power. Conversely, fusion methods excel at integrating information but often struggle to robustly differentiate between high-quality and noisy views, which can obscure final cluster boundaries and degrade performance. To address these complementary challenges, we propose GAPS (Granularity-Aware Pseudo Supervision), a novel MVC framework. GAPS introduces a granularity-aware supervision mechanism that generates a full hierarchy of pseudo-labels, enabling the selection of a supervision level that best aligns with the data's intrinsic multi-scale structure. Furthermore, to ensure a high-quality supervisory signal, it incorporates a reliability-aware view selection strategy using a novel Separation-Compactness Index (SCI) to identify and leverage the most informative view for pseudo-label generation. This dual approach ensures the supervisory signal is both structurally adaptive and derived from the most reliable source, leading to highly effective final representations. Extensive experiments on synthetic and real-world datasets demonstrate the effectiveness and superiority of GAPS over other competitors.

Abstract Recent advances in 3D Gaussian Splatting (3DGS) have achieved state‐of‐the‐art results for novel view synthesis. However, efficiently capturing high‐fidelity reconstructions of specific objects within complex scenes remains a significant challenge. A key limitation of existing active reconstruction methods is their reliance on scene‐level uncertainty metrics, which are often biased by irrelevant background clutter and lead to inefficient view selection for object‐centric tasks. We present OUGS, a novel framework that addresses this challenge with a more principled, physically‐grounded uncertainty formulation for 3DGS. Our core innovation is to derive uncertainty directly from the explicit physical parameters of the 3D Gaussian primitives (e.g., position, scale, rotation). By propagating the covariance of these parameters through the rendering Jacobian, we establish a highly interpretable uncertainty model. This foundation allows us to then seamlessly integrate semantic segmentation masks to produce a targeted, object‐aware uncertainty score that effectively disentangles the object from its environment. This allows for a more effective active view selection strategy that prioritizes views critical to improving object fidelity. Experimental evaluations on public datasets demonstrate that our approach significantly improves the efficiency of the 3DGS reconstruction process and achieves higher quality for targeted objects compared to existing state‐of‐the‐art methods, while also serving as a robust uncertainty estimator for the global scene.

Postural Instability Theory (PIT) proposes that individuals who are naturally unstable on their feet are more susceptible to cybersickness. We hypothesize that this relationship extends to locomotive VR, such that people who exhibit greater instability when walking without VR will also be more susceptible to cybersickness in a locomotive VR setup. To test this, we analyzed participants' natural walking kinematics alongside their cybersickness responses and kinematic patterns during mobile VR use. Our results showed that vertical Center of Mass movement during pre-VR walking showed promise for identifying individuals susceptible to cybersickness. Spatial stability metrics emerged as stronger predictors of cybersickness than time-series measures, suggesting that spatial characteristics of gait may be more informative indicators of susceptibility in mobile VR contexts. These findings highlight the importance of accounting for baseline postural stability when designing and personalizing mobile VR experiences.

Educator Perceptions of XRAuthor: An Accessible Tool for Authoring Learning Content with Different Immersion Levels

In ordinal classification, misclassifying neighboring ranks is common, yet the consequences of these errors are not the same. For example, misclassifying benign tumor categories is less consequential, compared to an error at the pre-cancerous to cancerous threshold, which could profoundly influence treatment choices. Despite this, existing ordinal classification methods do not account for the varying importance of these margins, treating all neighboring classes as equally significant. To address this limitation, we propose CLOC, a new margin-based contrastive learning method for ordinal classification that learns an ordered representation based on the optimization of multiple margins with a novel multi-margin n-pair loss (MMNP). CLOC enables flexible decision boundaries across key adjacent categories, facilitating smooth transitions between classes and reducing the risk of overfitting to biases present in the training data. We provide empirical discussion regarding the properties of MMNP and show experimental results on five real-world image datasets (Adience, Historical Colour Image Dating, Knee Osteoarthritis, Indian Diabetic Retinopathy Image, and Breast Carcinoma Subtyping) and one synthetic dataset simulating clinical decision bias. Our results demonstrate that CLOC outperforms existing ordinal classification methods and show the interpretability and controllability of CLOC in learning meaningful, ordered representations that align with clinical and practical needs.

Augmented Reality (AR) devices, emerging as prominent mobile interaction platforms, face challenges in user safety, particularly concerning oncoming vehicles. While some solutions leverage onboard camera arrays, these cameras often have limited field-of-view (FoV) with front or downward perspectives. Addressing this, we propose a new out-of-view semantic segmentation task and Segment Beyond View (SBV), a novel audio-visual semantic segmentation method. SBV supplements the visual modality, which miss the information beyond FoV, with the auditory information using a teacher-student distillation model (Omni2Ego). The model consists of a vision teacher utilising panoramic information, an auditory teacher with 8-channel audio, and an audio-visual student that takes views with limited FoV and binaural audio as input and produce semantic segmentation for objects outside FoV. SBV outperforms existing models in comparative evaluations and shows a consistent performance across varying FoV ranges and in monaural audio settings.

Vision-and-Language Navigation (VLN) task aims to enable AI agents to accurately understand and follow natural language instructions to navigate through real-world environments, ultimately reaching specific target locations. We recognise a promising opportunity to extend VLN to a comparable navigation task that holds substantial significance in our daily lives, albeit within the virtual realm: navigating websites on the Internet. This paper proposes a new task named Vision-and-Language Navigation on Websites (WebVLN), where we use question-based instructions to train an agent, emulating how users naturally browse websites. Unlike the existing VLN task that only pays attention to vision and instruction (language), the WebVLN agent further considers underlying web-specific content like HTML, which could not be seen on the rendered web pages yet contain rich visual and textual information. Toward this goal, we contribute a dataset, WebVLN-v1, and introduce a novel approach called Website-aware VLN Network (WebVLN-Net), which is built upon the foundation of state-of-the-art VLN techniques. Experimental results show that WebVLN-Net outperforms current VLN and web-related navigation methods. We believe that the introduction of the newWebVLN task and its dataset will establish a new dimension within the VLN domain and contribute to the broader vision-and-language research community. Code is available at: https://github.com/WebVLN/WebVLN.

Augmented reality (AR) headsets could provide useful information to users, but they may also be a source of distraction. Previous works have explored using AR to enhance pedestrian safety by providing real-time warnings, but there has been little research on the impact of modality matching between distractions and warnings on pedestrian street crossing behaviour. We conducted a VR experiment using a within-subjects 2-by-2 design (N=24) with four conditions: (auditory distraction, visual distraction) $\times$ (auditory warning, visual warning). When experienced conditions with mismatched modalities, participants exhibited more cautious street crossing behaviours, such as reduced walking speed, and increased scan range after receiving the warning, and significantly faster reaction times to the incoming vehicle. The participants also expressed a preference for warnings to be presented in a modality different from the distraction. Our findings suggest that in the context of utilizing AR for pedestrian road safety, future AR interfaces should incorporate a warning modality that differs from the one causing distraction.

Previous studies showed that natural walking reduces the susceptibility to VR sickness. However, many users still experience VR sickness when wearing VR headsets that allow free walking in room-scale spaces. This paper studies VR sickness and postural instability while the user walks in an immersive virtual environment using an electroencephalogram (EEG) headset and a full-body motion capture system. The experiment induced VR sickness by gradually increasing the translation gain beyond the user's detection threshold. A between-group comparison between participants with and without VR sickness symptoms found some significant differences in postural stability but found none on gait patterns during the walking. In the EEG analysis, the group with VR sickness showed a reduction of alpha power, a phenomenon previously linked to a higher workload and efforts to maintain postural control. In contrast, the group without VR sickness exhibited brain activities linked to fine cognitive-motor control. The EEG result provides new insights into the postural instability theory: participants with VR sickness could maintain their postural stability at the cost of a higher cognitive workload. Our result also indicates that the analysis of lower-frequency power could complement behavioral data for continuous VR sickness detection in both stationary and mobile VR setups.

Advances in virtual reality technology have greatly benefited the acrophobia research field. Virtual reality height exposure is a reliable method of inducing stress with low variance across ages and demographics. When creating a virtual height exposure environment, researchers have often used haptic feedback elements to improve the sense of realism of a virtual environment. While the quality of the rendered for the virtual environment increases over time, the physical environment is often simplified to a conservative passive haptic feedback platform. The impact of the increasing disparity between the virtual and physical environment on the induced stress levels is unclear. This article presents an experiment that explored the effect of combining an elevated physical platform with different levels of virtual heights to induce stress. Eighteen participants experienced four different conditions of varying physical and virtual heights. The measurements included gait parameters, heart rate, heart rate variability, and electrodermal activity. The results show that the added physical elevation at a low virtual height shifts the participant's walking behaviour and increases the perception of danger. However, the virtual environment still plays an essential role in manipulating height exposure and inducing physiological stress. Another finding is that a person's behaviour always corresponds to the more significant perceived threat, whether from the physical or virtual environment.

The COVID-19 pandemic has created the distinct challenge for the piloting of drones/other UAVs for researchers and educators who are restricted to working remotely. We propose a Remote Visual Line-of-Sight system that leverages the advantages of Virtual Reality (VR) and motion capture to allow users to fly a real-world drone from a remote location. The system was developed while our researcher (VR operator) was remotely working in Vietnam with the enclosed real-world environment located in Australia. Our paper will present the system design and the challenges found during the development of our system.

Room-scale 3D position tracking enables users to explore a virtual environment by physically walking, which improves comfort and the level of immersion. However, when users walk with their eyesight blocked by a head-mounted display, they may unexpectedly lose their balance and fall if they bump into real-world obstacles or unintentionally shift their center of mass outside the margin of stability. This paper evaluates balance recovery methods and intervention timing during the use of VR with the assumption that the onset of a fall is given. Our experiment followed the tether-release protocol during clinical research and induced a fall while a subject was engaged in a secondary 3D object selection task. The experiment employed a two-by-two design that evaluated two assistive techniques, i.e., video-see-through and auditory warning at two different timings, i.e., at fall onset and 500ms prior to fall onset. The data from 17 subjects showed that video-see-through triggered 500 ms before the onset of fall can effectively help users recover from falls. Surprisingly, video-see-through at fall onset has a significant negative impact on balance recovery and produces similar results to those of the baseline condition (no intervention).

Designing immersion is the key challenge in virtual reality; this challenge has driven advancements in displays, rendering and recently, haptics. To increase our sense of physical immersion, for instance, vibrotactile gloves render the sense of touching, while electrical muscle stimulation (EMS) renders forces. Unfortunately, the established metric to assess the effectiveness of haptic devices relies on the user's subjective interpretation of unspecific, yet standardized, questions.

Recent advances of the affordable virtual reality headsets make virtual reality training an economical choice when compared to traditional training. However, these virtual reality devices present a range of different levels of virtual reality fidelity and interactions. Few works have evaluated their validity against the traditional training formats. This paper presents a study that compares the learning efficiency of a bimanual gearbox assembly task among traditional training, virtual reality training with direct 3D inputs (HTC VIVE), and virtual reality training without 3D inputs (Google Cardboard). A pilot study was conducted and the result shows that HTC VIVE brings the best learning outcomes.

No abstract available.

We present TrussFab, an integrated end-to-end system that allows users to fabricate large scale structures that are sturdy enough to carry human weight. TrussFab achieves the large scale by complementing 3D print with plastic bottles. It does not use these bottles as "bricks" though, but as beams that form structurally sound node-link structures, also known as trusses, allowing it to handle the forces resulting from scale and load. TrussFab embodies the required engineering knowledge, allowing non-engineers to design such structures and to validate their design using integrated structural analysis. We have used TrussFab to design and fabricate tables and chairs, a 2.5 m long bridge strong enough to carry a human, a functional boat that seats two, and a 5 m diameter dome.

This paper proposes a 3D vision based object grasping posture learning system. In this system, the robot recognizes the orientation of the object to decide the grasping posture, whereas selects a feasible grasping point by detecting the surrounding. When the planned posture is not good enough, the proposed learning system adjusts the position of the end effector real time. The learning system is inspired by a book entitled, Thinking, Fast and Slow, and consists of two subsystems. The subsystem I judges whether the pose of the object is learned before, and plans a grasping posture by past experience. When the pose of the object is not learned before, the subsystem II learns a position adjustment by the real time information form the motor angels and the images. Finally, the method proposed in this paper is applied to the home service robot and is proven the feasibility by the experimental results.

Digital image editing is usually an iterative process; users repetitively perform short sequences of operations, as well as undo and redo using history navigation tools. In our collected data, undo, redo and navigation constitute about 9 percent of the total commands and consume a significant amount of user time. Unfortunately, such activities also tend to be tedious and frustrating, especially for complex projects.

Recently, researchers started to engineer not only the outer shape of objects, but also their internal microstructure. Such objects, typically based on 3D cell grids, are also known as metamaterials. Metamaterials have been used, for example, to create materials with soft and hard regions.

This paper presents Platener, a system that allows quickly fabricating intermediate design iterations of 3D models, a process also known as low-fidelity fabrication. Platener achieves its speed-up by extracting straight and curved plates from the 3D model and substituting them with laser cut parts of the same size and thickness. Only the regions that are of relevance to the current design iteration are executed as full-detail 3D prints. Platener connects the parts it has created by automatically inserting joints. To help fast assembly it engraves instructions. Platener allows users to customize substitution results by (1) specifying fidelity-speed tradeoffs, (2) choosing whether or not to convert curved surfaces to plates bent using heat, and (3) specifying the conversion of individual plates and joints interactively. Platener is designed to best preserve the fidelity of func-tional objects, such as casings and mechanical tools, all of which contain a large percentage of straight/rectilinear elements. Compared to other low-fab systems, such as faBrickator and WirePrint, Platener better preserves the stability and functionality of such objects: the resulting assemblies have fewer parts and the parts have the same size and thickness as in the 3D model. To validate our system, we converted 2.250 3D models downloaded from a 3D model site (Thingiverse). Platener achieves a speed-up of 10 or more for 39.5% of all objects.

Physical sketching of 3D wireframe models, using a hand-held plastic extruder, allows users to explore the design space of 3D models efficiently. Unfortunately, the scale of these devices limits users' design explorations to small-scale objects. We present protopiper, a computer aided, hand-held fabrication device, that allows users to sketch room-sized objects at actual scale. The key idea behind protopiper is that it forms adhesive tape into tubes as its main building material, rather than extruded plastic or photopolymer lines. Since the resulting tubes are hollow they offer excellent strength-to-weight ratio, thus scale well to large structures. Since the tape is pre-coated with adhesive it allows connecting tubes quickly, unlike extruded plastic that would require heating and cooling in the kilowatt range. We demonstrate protopiper's use through several demo objects, ranging from more constructive objects, such as furniture, to more decorative objects, such as statues. In our exploratory user study, 16 participants created objects based on their own ideas. They rated the device as being "useful for creative exploration", "its ability to sketch at actual scale helped judge fit", and "fun to use."

Laser cutters are useful for rapid prototyping because they are fast. However, they only produce planar 2D geometry. One approach to creating non-planar objects is to cut the object in horizontal slices and to stack and glue them. This approach, however, requires manual effort for the assembly and time for the glue to set, defeating the purpose of using a fast fabrication tool. We propose eliminating the assembly step with our system LaserStacker. The key idea is to use the laser cutter to not only cut but also to weld. Users place not one acrylic sheet, but a stack of acrylic sheets into their cutter. In a single process, LaserStacker cuts each individual layer to shape (through all layers above it), welds layers by melting material at their interface, and heals undesired cuts in higher layers. When users take out the object from the laser cutter, it is already assembled. To allow users to model stacked objects efficiently, we built an extension to a commercial 3D editor (SketchUp) that provides tools for defining which parts should be connected and which remain loose. When users hit the export button, LaserStacker converts the 3D model into cutting, welding, and healing instructions for the laser cutter. We show how LaserStacker does not only allow making static objects, such as architectural models, but also objects with moving parts and simple mechanisms, such as scissors, a simple pinball machine, and a mechanical toy with gears.

Abstract We presentfacetons, geometric modeling primitives designed for building architectural models especially effective for a virtual environment where six degrees of freedom input devices are available. Afacetonis an oriented point floating in the air and defines a plane of infinite extent passing through the point. The polygonal mesh model is constructed by taking the intersection of the planes associated with thefacetons. With the simple interaction offaceton, users can easily create 3D architecture models. Thefacetonprimitive and its interaction reduce the overhead associated with standard polygonal mesh modeling, where users have to manually specify vertexes and edges which could be far away. Thefacetonrepresentation is inspired by the research on boundary representations (B‐rep) and constructive solid geometry, but it is driven by a novel adaptive bounding algorithm and is specifically designed for 3D modeling activities in an immersive virtual environment. We describe the modeling method and our current implementation. The implementation is still experimental but shows potential as a viable alternative to traditional modeling methods. Copyright © 2014 John Wiley & Sons, Ltd.

3D modelers often wish to showcase their models for sharing or review purposes. This may consist of generating static viewpoints of the model or authoring animated fly-throughs. Manually creating such views is often tedious and few automatic methods are designed to interactively assist the modelers with the view authoring process. We present a view authoring assistance system that supports the creation of informative view points, view paths, and view surfaces, allowing modelers to author the interactive navigation experience of a model. The key concept of our implementation is to analyze the model's workflow history, to infer important regions of the model and representative viewpoints of those areas. An evaluation indicated that the viewpoints generated by our algorithm are comparable to those manually selected by the modeler. In addition, participants of a user study found our system easy to use and effective for authoring viewpoint summaries.

Painting is a major form of content creation, offering unlimited control and freedom of expression. However, it can involve tedious manual repetitions, such as stippling large regions or hatching complex contours. Thus, a central goal in digital painting research is to automate tedious repetitions while allowing user control. Existing methods impose a sequential order, in which a small exemplar is prepared and then cloned through additional gestures. Such sequential mode may break the continuous, spontaneous flow of painting. Moreover, it is more suitable for homogeneous areas than nuanced variations common in real paintings. We present an interactive digital painting system that auto-completes tedious repetitions while preserving nuanced variations and maintaining natural flows. Specifically, users paint as usual, while our system records and analyzes their workflows. When potential repetition is detected, our system predicts what the user might want to draw and offers auto-completes that adjust to the existing shape-color context. Our method eliminates the need for sequential creation-cloning and better adapts to the local painting contexts. Furthermore, users can choose to accept, ignore, or modify those predictions and thus maintain full control. Our method can be considered as the painting analogy of auto-completes in common typing and IDE systems. We demonstrate the quality and usability of our system through painting results and a pilot user study.

We present an interactive system for ink splattering, a form of abstract arts that artists drip or pour inks onto the canvas. The user interface and interactive methods are designed to be analogous to the artsitic techniques of ink splattering in real world so that digital artists can easily create the vibrant patterns of splattering ink, which are otherwise difficult to achieve in image editing software. The core of our system is a novel three-stage ink splattering framework that simulates the physical-based interaction of ink with different mediums including brush heads, air and paper. We implemented the physical engine using CUDA and the whole simulation process runs in real-time.

We present faceton, a geometric modeling primitive designed for building architectural models, using a six degrees of freedom (DoF) input device in a virtual environment (VE). A faceton is given as an oriented point floating in the air and defines a plane of infinite extent passing through the point. The polygonal mesh model is constructed by taking the intersection of the planes associated with the facetons. With the simple drag-and-drop and group interaction of faceton, users can easily create 3D architecture models in the VE. The faceton primitive and its interaction reduce the overhead associated with standard polygonal mesh modeling in VE, where users have to manually specify vertexes and edges which could be far away. The faceton representation is inspired by the research on boundary representations (B-rep) and constructive solid geometry (CSG), but it is driven by a novel adaptive bounding algorithm and is specifically designed for the 3D modeling activities in an immersive virtual environment.

Abstract This paper presents an interactive system for ink splattering, a form of abstract art that artists splat ink onto the canvas. The default input device of our system is a pressure‐sensitive 2D stylus, the most common sketching tool for digital artists, and we propose two interaction mode:ink‐flicking modeandink‐dripping mode, that are designed to be analogous to the artistic techniques of ink splattering in real world. The core of our ink splattering system is a novel three‐stage ink splattering framework that simulates the physics‐based interaction of ink with different mediums including brush heads, air and paper. We have implemented the physical engine in CUDA and the whole simulation process runs at interactive speed.

Revision control is a vital component of digital project management and has been widely deployed for text files. Binary files, on the other hand, have received relatively less attention. This can be inconvenient for graphics applications that use a significant amount of binary data, such as images, videos, meshes, and animations. Existing strategies such as storing whole files for individual revisions or simple binary deltas could consume significant storage and obscure vital semantic information. We present a nonlinear revision control system for images, designed with the common digital editing and sketching workflows in mind. We use DAG (directed acyclic graph) as the core structure, with DAG nodes representing editing operations and DAG edges the corresponding spatial, temporal and semantic relationships. We visualize our DAG in RevG (revision graph), which provides not only as a meaningful display of the revision history but also an intuitive interface for common revision control operations such as review, replay, diff, addition, branching, merging, and conflict resolving. Beyond revision control, our system also facilitates artistic creation processes in common image editing and digital painting workflows. We have built a prototype system upon GIMP, an open source image editor, and demonstrate its effectiveness through formative user study and comparisons with alternative revision control systems.