1,720,973 research outputs found
Unity-VRlines: Towards a Modular eXtended Reality Unity Flight Simulator
No full textComputer-aided flight simulation systems (CAFSS) make it possible to simulate flying an airplane using software and hardware. These simulations range from simple programs to intricate, full-motion simulators that often integrate physical feedback and visual clues to create realistic and affordable entertainment and pilot training. Nowadays, eXtended Reality (XR) paradigms have been integrated in CAFSS, to increase the immersiveness and realism of the experience, demonstrating positive cognitive learning effects in training procedures. However, no extensive results for simulator effectiveness are available to this date, considering the reach of such systems is limited by the costly hardware and unavailability of open-source software. For this reason, we here introduce Unity-VRlines, an open-source modular virtual reality flight simulator baseline, based on the Unity game engine and the SteamVR SDK, that can be deployed in any compatible VR device. The system components and software architecture enables developers to add new flight control instructions, alter aircraft parts, and change the surrounding environment
Flying in XR: Bridging Desktop Applications in eXtended Reality through Deep Learning
No full textThe expanding realm of eXtended Reality (XR) has witnessed a surge in 3D experiences across diverse domains, undergoing significant transformations to define novel experiences. However, such experiences are often built from scratch, as there is a lack of tools that support augmenting existing non-immersive interfaces, like Desktop games and simulators, directly into XR. Such shortage is particularly exacerbated in the case of Mixed Reality (MR). Motivated by this, we present a novel middleware, Flying In XR (FIXR), leveraging Deep Learning to visualize and interact with desktop application views into XR. To demonstrate the flexibility of such an approach, we applied FIXR to a commercial Desktop flight simulator, sup-porting an MR experience. It is worth noticing that FIXR could be adapted to communicate with any desktop software with a camera that moves along the depth axis, opening new paths to enable user experiences in XR for a wide spectrum of applications
Preserving Family Album Photos with the HoloLens 2
No full textFollowing the leap set by the introduction of digital photography and spread of social media, the custom of creating and nourishing a family photo album is steadily declining. The feeling of waving goodbye to an old technology may hence explain why collectors and museums are increasingly searching for such type of photographs. In such scenario, a barrier set to the preservation and analysis of the enormous source of information is the unavailability of a fine-grained digitization and cataloguing infrastructure. An opportunity to resolve such issue is offered by augmented reality (AR): we envision an AR-application capable of individuating and segmenting photos, while browsing through a family album. In addition to this digitization step, we also include a second one where the digitized images are analyzed, tagging them with meta-data pertaining their presumed socio-historical context and their date. The addition of such meta-data is key not only from a cataloguing point of view, but also from a conservation perspective. In fact, analog photos are more often preserved when some information pertaining their subject and their date is known. To this aim, we experiment the use of the HoloLens 2 along with artificial intelligence paradigms
Searching for cultural relationships through deep learning models
Full text linkFamily album photo collections may reveal historical insights regarding specific cultures and times. In most cases, such photos are scattered among private homes and only available on paper or photographic film, thus making their analysis very cumbersome. Their study may also become difficult because of the number of photos that such collections contain. It would be exceedingly long to manually verify the characteristics of more than a few hundred photos, considering that often no associated descriptions are available. This work falls in the described domain, addressing the problem of dating an image resorting to the analysis of an analog family album photo dataset, namely IMAGO, containing photos shot in the 20th century. Thanks to the IMAGO dataset, it was possible to apply different deep learning-based architectures to date images belonging to photo albums without needing any other sources of information. In addition, with the implementation of cross-dataset experiments, which also involved models previously presented in the literature, it was possible to observe temporal shifts which may be due to known intercultural influences. Concluding, deep learning models revealed their potential not only in terms of their performance but also in terms of their possible applications to intercultural research
Deep Armocromia: A Novel Dataset for Face Seasonal Color Analysis and Classification
No full textRecent advancements in Artificial Intelligence and Computer Vision, in particular Deep Learning (DL), have transformed the analysis of human faces, enabling different tasks, ranging from classification to synthesis. Despite these advancements, color analysis in face images remains underexplored, especially concerning well-defined datasets and frameworks tailored to specific methodologies such as Season Color Analysis or Armocromia. Armocromia combines qualitative and quantitative approaches to determine personal color palettes based on an individual’s skin, hair, and eye color; for this, it is vastly adopted in the fashion world. However, we found a lack of datasets to train DL models to automatically discriminate among these classes. To this date, we introduce Deep Armocromia, a novel dataset comprising labeled face images categorized according to Armocromia Flow Theory, with a strict annotation protocol. We conduct experiments to validate the effectiveness of DL models in discriminating among Armocromia classes optimized on Deep Armocromia. Results underscore the challenges inherent to Armocromia classification and highlight opportunities for advancing DL architectures and optimization methodologies
CreAIXR: Fostering Creativity with Generative AI in XR environments
No full textFostering creativity is paramount for cultivating innovative minds capable of addressing complex challenges. Modern technologies like eXtended Reality (XR) and Artificial Intelligence (AI) may nurture grounds supporting creative thinking by providing immersive and manipulable environments. An open research question is how such technologies may best lead to such a possible result. To help move one step closer to an answer, we present a portable XR platform, namely CreAIXR, where objects may be creatively defined and manipulated with AI paradigms. CreAIXR leverages web technologies, XR, and generative AI where creatives are immersed in a composable experience, allowing them to collaborate and customize an immersive environment through XR paradigms and generative AI. We here describe this system along with its validation through experiments carried out with a group of individuals having a background in the field of visual arts
Applying deep learning approaches to mixed quantitative-qualitative analyses
Full text linkWe here verify whether a quantitative approach, i.e., a deep learning-based one, may be used to synthesize a model apt to perform specific qualitative analyses. To this aim, we leverage a previous contribution, where we approached the concrete problem of implementing a socio-historical classification toolchain for a collection of vernacular photos. In such a work, after individuating a corpus of vernacular photographs we devised the process that follows. First, we resorted to existing socio-historical categories derived from previous qualitative studies. Secondly, we involved the people included in the photos in the annotation process of a subset of the corpus of data. We then fine-tuned and deployed existing deep learning models to classify the entire corpus of data. Finally, we compared the results obtained with our approach to the ones obtained by a socio-historian. We hence here focus on the relationship between quantitative and qualitative methods considering the specific case of socio-historical analyses
Rethinking Augmented Wine Recognition
No full textOverlaying rendered virtual annotations on top of the camera view of the real world requires intensive use of computer vision paradigms for object recognition and tracking. This involves computationally intensive tasks and the availability of large-scale databases of ref-erence images. In some domains, a lack of reference images may be particularly disruptive. For example, with wine bottles, labels may not be available because (a) periodically changed by the winery, and (b) specific bottles may belong to the long tail, making label retrieval difficult or even impossible. In the following, we present Augmented Wine Recognition (AWR), a system that does not require any reference image optimized to perform an augmented reading of wine labels
HOCTOPUS: An Open-Source Cross-Reality tool to Augment Live-Streaming Remote Classes
No full textExtended Reality (XR) has gained significant attention, offering novel pathways for learning based on the sharing of 3D content. Nevertheless, additional efforts should be spent to understand the advantages set forth by Cross-Reality (CR) scenarios, for example including Augmented Reality (AR) and Mixed Reality (MR) experiences in the context of remote live-streaming classes. In this process, a necessary step entails designing, implementing, and making educational tools available in an open-source format. To this aim, we present HOCTOPUS, a CR system designed to support live remote teaching. HOCTOPUS provides a cross-reality educational experience: an MR application for teachers and an AR mobile app for students. HOCTOPUS lets teachers visualize, manipulate, and share 3D objects in a live-streaming fashion, supporting bidirectional communication and interactions with all students. The MR component is developed for the Hololens 2, offering flexible manipulation capabilities, while the AR component runs on Android mobile devices, providing affordable interactive visualization and manipulation. The proposed tool may be used as a building block to enhance the learning process and is made available to the educational community for its use, assessment, and extension
- …
