Mobile augmented reality (Mobile AR) is gaining increasing attention from both academia and industry. Hardware-based AR App-based are the two dominant platforms for applications. However, hardware-based implementation known to be costly lacks flexibility, while one requires additional downloading installation in advance inconvenient cross-platform deployment. In comparison, Web-based (Web can provide a pervasive experience users thanks many successful deployments of Web as lightweight...

Dedicated device-based and app-based augmented reality (AR) solutions have inherent limitations regarding cross-platform, pervasive AR application provisioning. Web-based (web AR), a promising lightweight cross-platform approach to AR, is gaining increasing attention owing its extensive areas. However, for computationally intensive applications, the weak computational efficiency of current web browsers seriously hampers applications on large scale. The browser + cloud suffers from...

The popularity of wearable devices and smartphones has fueled the development Mobile Augmented Reality (MAR), which provides immersive experiences over real world using techniques, such as computer vision deep learning. However, hardware-specific MAR is costly heavy, App-based requires an additional download installation it also lacks cross-platform ability. These limitations hamper pervasive promotion MAR. This paper argues that mobile Web AR (MWAR) holds potential to become a practical...

Multi-user mobile Augmented Reality (AR) has been successfully used in various fields as a novel visual interaction technology. But current mainstream wearable device-based and app-based solutions are still facing cross-platform, real-time communication, intensive computing requirements. Mobile Web technology is envisioned to be promising supporting for cross-platform application of AR especially 5G networks, which provide pervasive communication resources thereby forming formidable...

The raw depth image captured by the indoor sen-sor usually has an extensive range of missing values due to inherent limitations such as inability perceive transparent objects and limited distance range. incomplete map burdens many downstream vision tasks, a rising number completion methods have been proposed alleviate this issue. While most existing meth-ods can generate accurate dense maps from sparse uniformly sampled maps, they are not suitable for complementing large contiguous regions...

Web-based Augmented Reality (Web AR) provides a lightweight, cross-platform, and pervasive AR solution. However, all of the current Web implementations still face some challenges, which greatly hinder promotion applications. Benefiting from Mobile Edge Computing (MEC) paradigm, in this paper, we propose MEC-based collaborative solution, can be regarded as feasible promising one. The edge server not only reduces network latency but also decreases bandwidth usage core networks. Prototype...

Web-based DNNs provide accurate object recognition to the mobile Web AR, which is newly emerging as a lightweight AR solution.Webbased are attracting great deal of attention.However, balancing UX against computing cost for DNN-based on difficult both self-contained and cloud-based offloading approaches, it latency-sensitive service but also has high requirements in terms networking abilities.Fortunately, 5G networks promise not only bandwidth latency improvement pervasive deployment edge...

Recently, following the rapid commercial deployment of 5G networks, next-generation mobile communication technology (6G) has been attracting increasing attention from global researchers and engineers. 6G is envisioned as a distributed, decentralized, intelligent innovative network. However, existing application provisioning still based on centralized service architecture, ubiquitous edge computing, decentralized AI technologies have not fully exploited. In this article, we analyze problems...

Deep learning technologies are empowering IoT devices with an increasing number of intelligent services. However, the contradiction between resource-constrained and intensive computing makes it common to transfer data cloud center for executing all DNN inference, or dynamically allocate computations center. Existing approaches perform a strong dependence on center, require support reliable stable network. Thus, may directly cause unreliable even unavailable service in extreme unstable...

Enabling deep learning technology on the mobile web can improve user's experience for achieving artificial intelligence in various fields. However, heavy DNN models and limited computing resources of are now unable to support executing computationally intensive DNNs when deploying a cloud platform. With help promising edge computing, we propose lightweight collaborative neural network web, named LcDNN, which contributes three aspects: (1) We design composite that reduces model size,...

Real-time holographic video communications enable immersive experiences for next-generation services in the future metaverse era. However, high-fidelity videos require high bandwidth and significant computation resources, which exceed transferring computing capacity of 5G networks. This article reviews state-of-the-art point cloud transmission techniques highlights critical challenges delivering such services. We further implement a preliminary prototype an AI-driven communication system...

Volumetric video provides a more immersive holographic virtual experience than conventional services such as 360-degree and reality (VR) videos. However, due to ultra-high bandwidth requirements, existing compression transmission technology cannot handle the delivery of real-time volumetric video. Unlike traditional methods approaches that extend streaming, we propose AITransfer, an AI-powered semantic-aware method for point cloud data (a popular format). AITransfer targets semantic-level...

In order to build a low-latency lightweight publish/subscribe (pub/sub) system for IOT services, we propose an efficient and scalable broker architecture, called Grid Quorum-based pub/sub (GQPS). As core component in the event-driven SOA framework this architecture organizes multiple brokers into quorum-based peer-to-peer topology topic searching. It also leverages searching algorithm caching strategy achieve small constant search latency. Lightweight RESTful interfaces make our GQPS more...

Based on advances in image processing technology and Web-enabling technologies for mobile devices, Augmented Reality (AR) Virtual (VR) has developed rapidly. The rendering interaction of 3D models is an important part AR VR applications closely related to user experience. However, since the existing WebGL JavaScript libraries Web-based (represented by three.js babylon.js) load entire model file at once, large-size with complex interactions cannot be rendered smoothly due limited data...

Network embedding (graph embedding) has become the focus of studying graph structure in recent years. In addition to research on homogeneous networks and heterogeneous networks, there are also some methods attempt solve problem dynamic network embedding. However, is no method specifically for networks. Therefore, this paper proposes a attention (HDGAN), which attempts use mechanism take heterogeneity dynamics into account at same time, so as better learn Our based three levels attention,...

Deep learning shows great promise in providing more intelligence to the mobile web, but insufficient infrastructure, heavy models, and intensive computation limit use of deep web applications. In this paper, we present DeepAdapter, a collaborative framework that ties with an edge server remote cloud allow executing on lower processing latency, energy, higher system throughput. DeepAdapter provides context-aware pruning algorithm incorporates network condition computing capability device fit...

Web-based Deep Neural Networks (DNNs) enhance the ability of object recognition and has attracted considerable attention in mobile Web AR other services. However, neither performing DNN inference on browsers locally nor offloading computations to cloud can strike a balance between accuracy efficiency; generally, rude methods are often accompanied by unsatisfactory accuracy. Collaborative approaches seem fill this gap coordinating distributed hierarchical computing resources, especially 5G...

Point cloud video provides a more immersive holographic virtual experience than conventional services such as 360 degree and reality (VR) video. However, the existing network bandwidth transmission technology can not carry real-time point streaming due to mass data volume, high processing overheads, extremely bandwidth-consuming. Unlike previous approaches that extend VR streaming, we propose AITransfer, an AI-powered bandwidth-aware adaptive technique driven by extracting transferring key...

Deep neural network (DNN) shows great promise in providing more intelligence to ubiquitous end devices. However, the existing partition-offloading schemes adopt data-parallel or model-parallel collaboration between devices and cloud, which does not make full use of resources for deep-level parallel execution. This article proposes eDDNN (i.e., enabling Distributed DNN), a collaborative inference scheme over heterogeneous using cross-platform Web technology, moving computation close devices,...

Recently, there has been a surge of the video services over Internet. However, service providers still have difficulties in providing high-quality streaming due to problem scheduling efficiency and wide fluctuations end-to-end delays existing multi-path algorithms. To solve these two problems affecting transmission quality, networks are expected capability dynamically managing network nodes for satisfying quality-of-service requirements, which is challenging issue media applications. Against...

Raw depth images captured in indoor scenarios frequently exhibit extensive missing values due to the inherent limitations of sensors and environments. For example, transparent materials elude detection by sensors; surfaces may introduce measurement inaccuracies their polished textures, extended distances, oblique incidence angles from sensor. The presence incomplete maps imposes significant challenges for subsequent vision applications, prompting development numerous completion techniques...