The N-back task is widely recognized as a paradigm for cognitive training due to its adaptability and effectiveness in mitigating working memory decline. However, its repetitive and monotonous design often leads to reduced engagement and poor long-term adherence, lowering intervention effectiveness. To address these limitations, researchers have turned to game-based interventions and immersive technologies to enhance user motivation and sustain training participation. In this study, we propose VRN-back, a music rhythm-based gamified cognitive training system in virtual reality (VR). The system combines the adaptive N-back paradigm with rhythm-game mechanics to increase enjoyment and motivation, while immersive interaction fosters engagement and enhances training outcomes. We conducted a seven-day study with 15 young adults to evaluate the system's impact on their working memory, transfer to related cognitive domains, and user experience. Results showed significant improvements in N-back accuracy, reaction time, and maximum N level, as well as transfer effects on attention and inhibitory control tasks. Subjective evaluations indicated good usability and user experience, with the System Usability Scale scoring 80.83/100, the short User Experience Questionnaire rating pragmatic quality as "Good" and hedonic quality as "Above Average", and the Pleasure-Arousal-Dominance scale confirming sustained positive emotional states. Simulator sickness remained minimal before and after training, ensuring usability. These findings suggest the feasibility and effectiveness of immersive rhythm-based gamified VR systems for cognitive training, laying a foundation for future research on long-term cognitive intervention technologies.

Flow‐field reconstruction from sparse sensor measurements remains a central challenge in modern fluid dynamics, as the need for high‐fidelity data often conflicts with practical limits on sensor deployment. Existing deep learning–based methods have demonstrated promising results, but they typically depend on simplifying assumptions such as two‐dimensional domains, predefined governing equations, synthetic datasets derived from idealized flow physics, and unconstrained sensor placement. In this work, we address these limitations by studying flow reconstruction under realistic conditions and introducing a \emph{directional transport‐aware Graph Neural Network (GNN)} that explicitly encodes both flow directionality and information transport. We further show that conventional sensor placement strategies frequently yield suboptimal configurations. To overcome this, we propose a novel \emph{Two‐Step Constrained PPO} procedure for Proximal Policy Optimization (PPO), which jointly optimizes sensor layouts by incorporating flow variability and accounts for reconstruction model's performance disparity with respect to sensor placement. We conduct comprehensive experiments under realistic assumptions to benchmark the performance of our reconstruction model and sensor placement policy. Together, they achieve significant improvements over existing methods.

Deep learning has shown promising results on dense three-dimensional (3D) model reconstruction from RGB images in recent years. However, the reconstruction of large-scale 3D models required for digital city remains very challenging even for such deep learning based methods. In this paper, we propose a convolutional neural network (CNN)-based Multi-View-Stereo (MVS) method that uses a double U-Net approach searching for image features. The proposed network first utilizes a double U-Net to extract the image features of a coarser resolution for the sake of reduced memory requirements. After that, the cost volume is built via the differentiable homography warping. The cascade structure is designed to extract the information in a small-scale cost volume before a large-scale cost volume carries out fusion and finer depth map estimation. As a result, the proposed network can efficiently produce highly accurate 3D point clouds using a fraction of the GPU memory and runtime required by conventional methods. Extensive experiments on the DTU benchmarks as well as the Tanks and Temples benchmarks confirm that the proposed network can achieve outstanding reconstruction accuracy and model completeness.

Accurate three-dimensional (3D) city models are critical for applications, such as digital city, urban planning and geo-graphic surveying and mapping. The synergy of deep learning and LiDAR-based Simultaneous Localization and Mapping (SLAM) has achieved some success in obtaining dense 3D color maps for large-scale outdoor scenes. In such a frame-work, this paper proposes a sensor-fusion system composed of a solid-state LiDAR, an inertial measurement unit (IMU) and a monocular camera by introducing image-guided depth completion into LiDAR SLAM. In the proposed system, the LiDAR-IMU odometry constructs the geometry structure of 3D maps, and subsequently, the images are used to render the texture of 3D color maps and guide the depth completion. In particular, the proposed system generates denser point clouds by exploiting the solid-state LiDAR before further increasing the density of point clouds via the depth completion. Our ex-perimental results show that the proposed system can achieve impressive 3D color models for large-scale outdoor scenes through faster scan and reconstruction process.

Abstract Large-scale high-resolution three-dimensional (3D) maps play a vital role in the development of smart cities. In this work, a novel deep learning-based multi-view-stereo method is proposed for reconstructing the 3D maps in large-scale urban environments by exploiting a monocular camera. Compared with other existing works, the proposed method can perform 3D depth estimation more efficiently in terms of computational complexity and graphics processing unit memory usage. As a result, the proposed method can practically perform depth estimation for each pixel before generating 3D maps for even large-scale scenes. Extensive experiments on the well-known DTU dataset and real-life data collected on our campus confirm the good performance of the proposed method.

Surface nodularity is an important image biomarker for cirrhosis. In this study, we explore how the location of the liver boundary curve may influence the liver surface nodularity (LSN). Based on computed tomography images of 7 patients with a normal liver and 9 patients with cirrhosis, we quantitatively estimate the LSN of the boundary curves selected from different locations of the liver. By repeating the estimation for different boundary curves, the difference in the variation of the LSN between the normal liver and cirrhosis is investigated. Receiver operating characteristic (ROC) analysis is used to assess the classification performance between the normal and cirrhotic livers. Results show that the LSN of the normal liver is significantly different from that of the cirrhosis, that the variation pattern of the LSN is different if the boundary curve is selected from different locations of the liver, and that the LSN leads to accurate classification between the normal and cirrhotic livers with the area under the ROC curve of 0.97- 0.98. We conclude that the LSN can be used to classify cirrhosis from the normal liver and studies are on-going to validate the conclusion.