RayGauss has achieved state-of-the-art rendering quality for novel-view synthesis on synthetic and indoor scenes by representing radiance and density fields with irregularly distributed elliptical basis functions, rendered via volume ray casting using a Bounding Volume Hierarchy (BVH). However, its computational cost prevents real-time rendering on real-world scenes. Our approach, RayGaussX, builds on RayGauss by introducing key contributions that accelerate both training and inference. Specifically, we incorporate volumetric rendering acceleration strategies such as empty-space skipping and adaptive sampling, enhance ray coherence, and introduce scale regularization to reduce false-positive intersections. Additionally, we propose a new densification criterion that improves density distribution in distant regions, leading to enhanced graphical quality on larger scenes. As a result, RayGaussX achieves 5x to 12x faster training and 50x to 80x higher rendering speeds (FPS) on real-world datasets while improving visual quality by up to +0.56 dB in PSNR. Project page with videos and code: https://raygaussx.github.io/.

Differentiable volumetric rendering-based methods made significant progress in novel view synthesis. On one hand, innovative methods have replaced the Neural Radiance Fields (NeRF) network with locally parame-terized structures, enabling high-quality renderings in a reasonable time. On the other hand, approaches have used differentiable splatting instead of NeRF's ray casting to optimize radiance fields rapidly using Gaussian kernels, allowing for fine adaptation to the scene. However, differentiable ray casting of irregularly spaced kernels has been scarcely explored, while splatting, despite enabling fast rendering times, is susceptible to clearly visible artifacts. Our work closes this gap by providing a physically consistent formulation of the emitted radiance c and density σ, decomposed with Gaussian functions associated with Spherical Gaussians/Harmonics for all-frequency colorimetric representation. We also introduce a method enabling differentiable ray casting of irregularly distributed Gaussians using an algorithm that integrates radiance fields slab by slab and leverages a BVH structure. This allows our approach to finely adapt to the scene while avoiding splatting artifacts. As a result, we achieve superior rendering quality compared to the state-of-the-art while maintaining reasonable training times and achieving inference speeds of 25 FPS on the Blender dataset. Project page with videos and code: raygauss.github.io

This article proposes a novel indoor magnetic field-based place recognition algorithm that is accurate and fast to compute. For that, we modified the generalized “Hough Transform” to process magnetic data (MagHT). It takes as input a sequence of magnetic measures whose relative positions are recovered by an odometry system and recognizes the places in the magnetic map where they were acquired. It also returns the global transformation from the coordinate frame of the input magnetic data to the magnetic map reference frame. Experimental results on several real datasets in large indoor environments demonstrate that the obtained localization error, recall, and precision are similar to or are better than state-of-the-art methods while improving the runtime by several orders of magnitude. Moreover, unlike magnetic sequence matching-based solutions such as DTW, our approach is independent of the path taken during the magnetic map creation.

In this work, we examine how Augmented Reality (AR) impacts user's situation awareness (SA) on elements secondary to an AR-assisted main task, i.e. not directly concerned by the main task. These secondary elements can still provide relevant information that we do not want the user to miss. A good understanding of user's awareness about them is therefore interesting, especially in a context of a daily use of AR, in which not all elements of user's environment are controlled. In this regard, we measured SA about secondary elements in an industrial workshop where the AR-assisted main task is a pedestrian navigation. We compared SA between three navigation guidance conditions: a paper map, a virtual path, and a virtual path with virtual cues about secondary elements. These secondary elements were either hazardous areas, for example, for mandatory helmets, or items which could be on user's path, for example, misplaced carts, boxes… We adapted an existing SA method evaluation to a real-world environment. With this method, participants were queried about their SA on three levels: perception, comprehension and projection about different items. We found that the use of AR decreased user's SA about secondary elements, and that this degradation mainly occurs at the perception level: with AR, participants are less likely to detect secondary elements. Participants still felt the most secure with AR and virtual cues about secondary elements.

Creating lifelike virtual humans for interactive virtual reality is a difficult task. Most current solutions rely either on crafting synthetic character models and animations, or on capturing real people with complex camera setups. As an alternative, we propose leveraging efficient learning-based models for human mesh estimation, and applying them to the popular form of immersive content that is 360° video. We demonstrate an implementation of this approach using available pre-trained models, and present user study results that show that the virtual agents generated with this method can be made more compelling by the use of idle animations and reactive verbal and gaze behavior.

Driver interaction with increasingly automated vehicles requires prior knowledge of system capabilities, operational know-how to use novel car equipment and responsiveness to unpredictable situations. With the purpose of getting drivers ready for autonomous driving, in a between-subject study sixty inexperienced participants were trained with an on-board video tutorial, an Augmented Reality (AR) program and a Virtual Reality (VR) simulator. To evaluate the transfer of training to real driving scenarios, a test drive on public roads was conducted implementing, for the first time in these conditions, the Wizard of Oz (WoZ) protocol. Results suggest that VR and AR training can foster knowledge acquisition and improve reaction time performance in take-over requests. Moreover, participants' behavior during the test drive highlights the ecological validity of the experiment thanks to the effective implementation of the WoZ methodology.

This work addresses the problem of using real-world data captured from a single viewpoint by a low-cost 360-degree camera to create an immersive and interactive virtual reality scene. We combine different existing state-of-the-art data enhancement methods based on pre-trained deep learning models to quickly and automatically obtain 3D scenes with animated character models from a 360-degree video. We provide details on our implementation and insight on how to adapt existing methods to 360-degree inputs. We also present the results of a user study assessing the extent to which virtual agents generated by this process are perceived as present and engaging.

In conditionally automated vehicles, drivers can engage in secondary activities while traveling to their destination. However, drivers are required to appropriately respond, in a limited amount of time, to a take-over request when the system reaches its functional boundaries. In this context, Virtual Reality systems represent a promising training and learning tool to properly familiarize drivers with the automated vehicle and allow them to interact with the novel equipment involved. In this study, the effectiveness of an Head-Mounted display (HMD)-based training program for acquiring interaction skills in automated cars was compared to a user manual and a fixed-base simulator. Results show that the training system affects the take-over performances evaluated in a test drive in a high-end driving simulator. Moreover, self-reported measures indicate that the HMD-based training is preferred with respect to the other systems.

This paper focuses on the acceptability of human-robot collaboration in industrial environments. A use case was designed in which an operator and a robot had to work side-by-side on automotive assembly lines, with different levels of co-presence. This use case was implemented both in a physical and in a virtual situation using virtual reality. A user study was conducted with operators from the automotive industry. The operators were asked to assess the acceptability to work side-by-side with the robot through questionnaires, and physiological measures (heart rate and skin conductance) were taken during the user study. The results showed that working close to the robot imposed more constraints on the operators and required them to adapt to the robot. Moreover, an increase in skin conductance level was observed after working close to the robot. Although no significant difference was found in the questionnaires results between the physical and virtual situations, the increase in physiological measures was significant only in the physical situation. This suggests that virtual reality may be a good tool to assess the acceptability of human-robot collaboration and draw preliminary results through questionnaires, but that physical experiments are still necessary to a complete study, especially when dealing with physiological measures.

Robots are becoming more and more present in our everyday life: they are already used for domestic tasks, for companionship activities, and soon they will be used to assist humans and collaborate with them in their work. Human-robot collaboration has already been studied in the industry, for ergonomics and efficiency purposes, but more from a safety than from an acceptability point of view. In this work, we focused on how people perceive robots in a collaboration task and we proposed to use virtual reality as a simulation environment to test different parameters, by making users collaborate with virtual robots. A simple use case was implemented to compare different robot appearances and different robot movements. Questionnaires and physiological measures were used to assess the acceptability level of each condition with a user study. The results showed that the perception of robot movements depended on robot appearance and that a more anthropomorphic robot, both in its appearance and movements, was not necessarily better accepted by the users in a collaboration task. Finally, this preliminary use case was also the opportunity to guarantee the relevance of using such a methodology --- based on virtual reality, questionnaires and physiological measures --- for future studies.

In this paper, we introduce a new deformation method adapted to immersive design. The use of Virtual Reality (VR) in the design process implies a physical displacement of project actors and data between the virtual reality facilities and the design office. The decisions taken in the immersive environment are manually reflected on the Computed Aided Design (CAD) system. This increases the design time and breaks the continuity of data workflow. On this basis, there is a clear demand among the industry for tools adapted to immersive design. But few methods exist that encompass CAD problematic in VR. For this purpose, we propose a new method, called D3, for Draw, Deform and Design, based on a 2 step manipulation paradigm, consisting with 1) area selection and 2) path drawing, and a final refining and fitting phase. Our method is discussed on the basis of a set of CAD deformation scenarios.

This poster presents a novel approach for integrating, as insets, high resolution stereoscopic areas within a wider low resolution stereoscopic image. The objective is to propose ways to increase local resolution in virtual reality systems, with the use of a small number of additional projectors. Our novel approach consists of inserting high resolution areas whose shape depends on the geometry of 3D objects. For comparison purposes, an existing method is implemented, it consists of setting a rectangular high resolution inset. Both methods are implemented in a large screen VR system with active stereoscopy and head tracking. A user study meant to check for shape perception discrepancies between the two methods was performed. Results show that the novel technique does not introduce any perception distortion in shape size.