High-quality web-based volume rendering in real-time

Author(s):  
Kongyot Wangkaoom ◽  
Paruj Ratanaworabhan ◽  
Saowapak S. Thongvigitmanee
2014 ◽  
Vol 5 (3) ◽  
pp. 1
Author(s):  
Márcio Cerqueira de Farias Macedo ◽  
Antônio Lopes Apolinário Júnior ◽  
Antonio Carlos dos Santos Souza ◽  
Gilson Antônio Giraldi

To provide on-patient medical data visualization, a medical augmented reality environment must support volume rendering, accurate tracking, real-time performance and high visual quality in the final rendering. Another interesting feature is markerless registration, to solve the intrusiveness introduced by the use of fiducial markers for tracking. In this paper we address the problem of on-patient medical data visualization in a real-time high-quality markerless augmented reality environment. The medical data consists of a volume reconstructed from 3D computed tomography image data. Markerless registration is done by generating a 3D reference model of the region of interest in the patient and tracking it from the depth stream of an RGB-D sensor. From the estimated camera pose, the volumetric medical data and the reference model are combined allowing a visualization of the patient as well as part of his anatomy. To improve the visual perception of the scene, focus+context visualization is used in the augmented reality scene to dynamically define which parts of the medical volume will be visualized in the context of the patient’s image. Moreover, context-preserving volume rendering is employed to dynamically control which parts of the volume will be rendered. The results obtained show that the markerless environment runs in real-time and the techniques applied greatly improve the visual quality of the final rendering.


2020 ◽  
Vol 2020 (14) ◽  
pp. 378-1-378-7
Author(s):  
Tyler Nuanes ◽  
Matt Elsey ◽  
Radek Grzeszczuk ◽  
John Paul Shen

We present a high-quality sky segmentation model for depth refinement and investigate residual architecture performance to inform optimally shrinking the network. We describe a model that runs in near real-time on mobile device, present a new, highquality dataset, and detail a unique weighing to trade off false positives and false negatives in binary classifiers. We show how the optimizations improve bokeh rendering by correcting stereo depth misprediction in sky regions. We detail techniques used to preserve edges, reject false positives, and ensure generalization to the diversity of sky scenes. Finally, we present a compact model and compare performance of four popular residual architectures (ShuffleNet, MobileNetV2, Resnet-101, and Resnet-34-like) at constant computational cost.


2021 ◽  
Vol 2021 (2) ◽  
Author(s):  
Qiang Chen ◽  
Jianyuan Xiao ◽  
Peifeng Fan

Abstract A class of high-order canonical symplectic structure-preserving geometric algorithms are developed for high-quality simulations of the quantized Dirac-Maxwell theory based strong-field quantum electrodynamics (SFQED) and relativistic quantum plasmas (RQP) phenomena. With minimal coupling, the Lagrangian density of an interacting bispinor-gauge fields theory is constructed in a conjugate real fields form. The canonical symplectic form and canonical equations of this field theory are obtained by the general Hamilton’s principle on cotangent bundle. Based on discrete exterior calculus, the gauge field components are discreted to form a cochain complex, and the bispinor components are naturally discreted on a staggered dual lattice as combinations of differential forms. With pull-back and push-forward gauge covariant derivatives, the discrete action is gauge invariant. A well-defined discrete canonical Poisson bracket generates a semi-discrete lattice canonical field theory (LCFT), which admits the canonical symplectic form, unitary property, gauge symmetry and discrete Poincaré subgroup, which are good approximations of the original continuous geometric structures. The Hamiltonian splitting method, Cayley transformation and symmetric composition technique are introduced to construct a class of high-order numerical schemes for the semi-discrete LCFT. These schemes involve two degenerate fermion flavors and are locally unconditional stable, which also preserve the geometric structures. Admitting Nielsen-Ninomiya theorem, the continuous chiral symmetry is partially broken on the lattice. As an extension, a pair of discrete chiral operators are introduced to reconstruct the lattice chirality. Equipped with statistically quantization-equivalent ensemble models of the Dirac vacuum and non-trivial plasma backgrounds, the schemes are expected to have excellent performance in secular simulations of relativistic quantum effects, where the numerical errors of conserved quantities are well bounded by very small values without coherent accumulation. The algorithms are verified in detail by numerical energy spectra. Real-time LCFT simulations are successfully implemented for the nonlinear Schwinger mechanism induced e-e+ pairs creation and vacuum Kerr effect, where the nonlinear and non-perturbative features captured by the solutions provide a complete strong-field physical picture in a very wide range, which open a new door toward high-quality simulations in SFQED and RQP fields.


Sensors ◽  
2021 ◽  
Vol 21 (12) ◽  
pp. 4045
Author(s):  
Alessandro Sassu ◽  
Jose Francisco Saenz-Cogollo ◽  
Maurizio Agelli

Edge computing is the best approach for meeting the exponential demand and the real-time requirements of many video analytics applications. Since most of the recent advances regarding the extraction of information from images and video rely on computation heavy deep learning algorithms, there is a growing need for solutions that allow the deployment and use of new models on scalable and flexible edge architectures. In this work, we present Deep-Framework, a novel open source framework for developing edge-oriented real-time video analytics applications based on deep learning. Deep-Framework has a scalable multi-stream architecture based on Docker and abstracts away from the user the complexity of cluster configuration, orchestration of services, and GPU resources allocation. It provides Python interfaces for integrating deep learning models developed with the most popular frameworks and also provides high-level APIs based on standard HTTP and WebRTC interfaces for consuming the extracted video data on clients running on browsers or any other web-based platform.


Sign in / Sign up

Export Citation Format

Share Document