Virtual Reality Intelligent Hardware最新文献

Background

The uncanny valley hypothesis states that users may experience discomfort when interacting with almost human-like artificial characters. Advancements in artificial intelligence, robotics, and computer graphics have led to the development of life-like virtual humans and humanoid robots. Revisiting this hypothesis is necessary to check whether they positively or negatively affect the current population, who are highly accustomed to the latest technologies.

Methods

In this study, we present a unique evaluation of the uncanny valley hypothesis by allowing participants to interact live with four humanoid robots that have varying levels of human-likeness. Each participant completed a survey questionnaire to evaluate the affinity of each robot. Additionally, we used deep learning methods to quantify the participants’ emotional states using multimodal cues, including visual, audio, and text cues, by recording the participant–robot interactions.

Results

Multi-modal analysis and surveys provided interesting results and insights into the uncanny valley hypothesis.

Background

Exploring correspondences across multiview images is the basis of various computer vision tasks. However, most existing methods have limited accuracy under challenging conditions.

Method

To learn more robust and accurate correspondences, we propose DSD-MatchingNet for local feature matching in this study. First, we develop a deformable feature extraction module to obtain multilevel feature maps, which harvest contextual information from dynamic receptive fields. The dynamic receptive fields provided by the deformable convolution network ensure that our method obtains dense and robust correspondence. Second, we utilize sparse-to-dense matching with symmetry of correspondence to implement accurate pixel-level matching, which enables our method to produce more accurate correspondences.

Result

Experiments show that our proposed DSD-MatchingNet achieves a better performance on the image matching benchmark, as well as on the visual localization benchmark. Specifically, our method achieved 91.3% mean matching accuracy on the HPatches dataset and 99.3% visual localization recalls on the Aachen Day-Night dataset.

Background

Virtual reality (VR) applications can be used to provide comprehensive training scenarios that are difficult or impossible to represent in physical configurations. This includes team training for emergency services such as firefighting. Creating a high level of immersion is essential for achieving effective virtual training. In this respect, motion-capture systems offer the possibility of creating highly immersive multi-user training experiences, including full-body avatars.

Methods

This study presents a preliminary prototype that helps extinguish a virtual fire on a container ship as a VR training scenario. The prototype provides a full-body and multi-user VR experience based on the synthesis of position data provided by the motion-capture system and orientation data from the VR headsets. Moreover, the prototype facilitates an initial evaluation of the results.

Results

The results confirm the value of using VR for training procedures that are difficult to implement in the real world. Furthermore, the results show that motion-capture-based VR technologies are particularly useful for firefighting training, in which participants can collaborate in difficult-to-access environments. However, this study also indicates that increasing the immersion in such training remains a challenge.

Conclusions

This study presents a prototypical VR application that enables the multi-user training of maritime firefighters. Future research should evaluate the initial results, provide more extensive training scenarios, and measure the training progress.

Background

Data-driven event analysis has gradually become the backbone of modern competitive sports analysis. Competitive sports data analysis tasks increasingly use computer vision and machine-learning models for intelligent data analysis. Existing sports visualization systems focus on the player–team data visualization, which is not intuitive enough for team season win–loss data and game time-series data visualization and neglects the prediction of all-star players.

Methods

This study used an interactive visualization system designed with parallel aggregated ordered hypergraph dynamic hypergraphs, Calliope visualization data story technology, and iStoryline narrative visualization technology to visualize the regular statistics and game time data of players and teams. NPIPVis includes dynamic hypergraphs of a teamʹs wins and losses and game plot narrative visualization components. In addition, an integrated learning-based all-star player prediction model, SRR-voting, which starts from the existing minority and majority samples, was proposed using the synthetic minority oversampling technique and RandomUnderSampler methods to generate and eliminate samples of a certain size to balance the number of allstar and average players in the datasets. Next, a random forest algorithm was introduced to extract and construct the features of players and combined with the voting integrated model to predict the all-star players, using Grid- SearchCV, to optimize the hyperparameters of each model in integrated learning and then combined with five-fold cross-validation to improve the generalization ability of the model. Finally, the SHapley Additive exPlanations (SHAP) model was introduced to enhance the interpretability of the model.

Results

The experimental results of comparing the SRR-voting model with six common models show that the accuracy, F1-score, and recall metrics are significantly improved, which verifies the effectiveness and practicality of the SRR-voting model.

Conclusions

This study combines data visualization and machine learning to design a National Basketball Association data visualization system to help the general audience visualize game data and predict all-star players; this can also be extended to other sports events or related fields.

Background

Gesture drawing is a type of fluid, fast sketch with loose and roughly drawn lines that captures the motion and feeling of a subject. Although style transfer methods, which are able to learn a style from an input image and apply it to a secondary image, can reproduce many styles, they are currently unable to produce the flowing strokes of gesture drawings.

Method

In this paper, we present a method for producing gesture drawings that roughly depict objects or scenes with loose dancing contours and frantic textures. By following a gradient field, our method adapts stroke-based painterly rendering algorithms to produce long curved strokes. A rough, overdrawn appearance is created through a progressive refinement. In addition, we produce rough hatch strokes by altering the stroke direction. These add optional shading to gesture drawings.

Results

A wealth of parameters provide users the ability to adjust the output style, from short and rapid strokes to long and fluid strokes, and from swirling to straight lines. Potential stylistic outputs include pen-and-ink and colored pencil. We present several generated gesture drawings and discuss the application of our method to video.

Conclusion

Our stroke-based rendering algorithm produces convincing gesture drawings with numerous controllable parameters, permitting the creation of a variety of styles.

Background

Monocular depth estimation aims to predict a dense depth map from a single RGB image, and has important applications in 3D reconstruction, automatic driving, and augmented reality. However, existing methods directly feed the original RGB image into the model to extract depth features without avoiding the interference of depth-irrelevant information on depth-estimation accuracy, which leads to inferior performance.

Methods

To remove the influence of depth-irrelevant information and improve the depth-prediction accuracy, we propose RADepthNet, a novel reflectance-guided network that fuses boundary features. Specifically, our method predicts depth maps using the following three steps: (1) Intrinsic Image Decomposition. We propose a reflectance extraction module consisting of an encoder-decoder structure to extract the depth-related reflectance. Through an ablation study, we demonstrate that the module can reduce the influence of illumination on depth estimation. (2) Boundary Detection. A boundary extraction module, consisting of an encoder, refinement block, and upsample block, was proposed to better predict the depth at object boundaries utilizing gradient constraints. (3) Depth Prediction Module. We use an encoder different from (2) to obtain depth features from the reflectance map and fuse boundary features to predict depth. In addition, we proposed FIFADataset, a depth-estimation dataset applied in soccer scenarios.

Results

Extensive experiments on a public dataset and our proposed FIFADataset show that our method achieves state-of-the-art performance.

Background

Digital twins are virtual representations of devices and processes that capture the physical properties of the environment and operational algorithms/techniques in the context of medical devices and technologies. Digital twins may allow healthcare organizations to determine methods of improving medical processes, enhancing patient experience, lowering operating expenses, and extending the value of care. During the present COVID-19 pandemic, various medical devices, such as X-rays and CT scan machines and processes, are constantly being used to collect and analyze medical images. When collecting and processing an extensive volume of data in the form of images, machines and processes sometimes suffer from system failures, creating critical issues for hospitals and patients.

Methods

To address this, we introduce a digital-twin-based smart healthcare system integrated with medical devices to collect information regarding the current health condition, configuration, and maintenance history of the device/machine/system. Furthermore, medical images, that is, X-rays, are analyzed by using a deep-learning model to detect the infection of COVID-19. The designed system is based on the cascade recurrent convolution neural network (RCNN) architecture. In this architecture, the detector stages are deeper and more sequentially selective against small and close false positives. This architecture is a multi-stage extension of the RCNN model and sequentially trained using the output of one stage for training the other. At each stage, the bounding boxes are adjusted to locate a suitable value of the nearest false positives during the training of the different stages. In this manner, the arrangement of detectors is adjusted to increase the intersection over union, overcoming the problem of overfitting. We train the model by using X-ray images as the model was previously trained on another dataset.

Results

The developed system achieves good accuracy during the detection phase of COVID-19. The experimental outcomes reveal the efficiency of the detection architecture, which yields a mean average precision rate of 0.94.

This work surveys and illustrates multiple open challenges in the field of industrial Internet of Things (IoT)-based big data management and analysis in cloud environments. Challenges arising from the fields of machine learning in cloud infrastructures, artificial intelligence techniques for big data analytics in cloud environments, and federated learning cloud systems are elucidated. Additionally, reinforcement learning, which is a novel technique that allows large cloud-based data centers, to allocate more energy-efficient resources is examined. Moreover, we propose an architecture that attempts to combine the features offered by several cloud providers to achieve an energy-efficient industrial IoT-based big data management framework (EEIBDM) established outside of every user in the cloud. IoT data can be integrated with techniques such as reinforcement and federated learning to achieve a digital twin scenario for the virtual representation of industrial IoT-based big data of machines and room temperatures. Furthermore, we propose an algorithm for determining the energy consumption of the infrastructure by evaluating the EEIBDM framework. Finally, future directions for the expansion of this research are discussed.

Background

Digital twins offer rich potential for exploration in virtual reality (VR). Using interactive molecular simulation approaches, they enable a human operator to access the physical properties of molecular objects and to build, manipulate, and study their assemblies. Integrative modeling and drug design are important applications of this technology.

Methods

In this study, head-mounted virtual reality displays connected to molecular simulation engines were used to create interactive and immersive digital twins. They were used to perform tasks relevant to specific use cases.

Results

Three areas were investigated, including model building, rational design, and tangible models. Here, we report several membrane-embedded systems of ion channels, viral components, and artificial water channels. We were able to improve and create molecular designs based on digital twins.

Conclusions

The molecular application domain offers great opportunities, and most of the technical and technological aspects have been solved. Wider adoption is expected once the onboarding of VR is simplified and the technology gains wider acceptance.