China
Africa
Explore chapters and articles related to this topic
Access Networks Evolution
Published in Zhensheng Jia, Luis Alberto Campos, Coherent Optics for Access Networks, 2019
Luis Alberto Campos, Zhensheng Jia
Video-intensive technologies require the most bandwidth, and immersive applications leveraging Virtual Reality/Augmented Reality (VR/AR) are the most demanding out of all video applications. Current VR applications are little more than 360° video/panoramas. A low-quality 360° video requires at least a 30 Mbps connection, HD quality streams easily surpass 100 Mbps, and retina quality (4k+) streams approach Gbps territory. Looking one step ahead in the evolution of immersive videos, holographic displays or light-field displays are being prototyped. As the name indicates, light-field displays reproduce the original light fields through a display, generating the original electromagnetic waves that would have been generated from actual objects from a display window. This allows the viewer to move and change viewing angles and perceive the natural changes of the image, such as seeing objects behind the foreground appear with viewing angle changes. Light-field displays provide true 3D and immersive visual representation without headsets. It is estimated that a commercial size display would require a capacity around 1.5 Gbps [3].
Review on Pupil Segmentation Using CNN-Region of Interest
Published in Kamal Kumar Sharma, Akhil Gupta, Bandana Sharma, Suman Lata Tripathi, Intelligent Communication and Automation Systems, 2021
A. Swathi, Aarti, Sandeep Kumar
Applications such as border control, identity-based systems, biometric systems, iris-recognition systems, computer graphics, security-based systems, etc. are highly dependent upon segmentation. With increasing challenges in security issues, the iris is going to be a lead character in many applications. The pupil, as part of iris, is the basic and most tangible part to treat and dependable criteria for many of the applications that were developed on iris [1, 2]. Applications range from gaze, movement and drowsiness detection to age prediction. Therefore segmentation of the pupil is the source of finding and extracting information about a person. This kind of information is also used in psychology and cognitive science applications. It has applications in virtual reality too based upon gaze [3]. Machine learning is a trending technology that has already come into our lives with many challenges. It's been already proven for its efficiency and learning rates with many stand-alone algorithms. Convolutional neural networks (CNNs) are sure to be popular for the same reason. Combining CNN efficiency with graphics processing unit (GPU) based systems, the training and testing rate of CNNs became reliable in developing the machine learning based applications. Fully connected CNNs started showing a lot of improvement in training with data loss and accuracy loss values. Another approach is semantic segmentation, which is capable of identifying 50 000 objects at a time – a real value of using CNN. These methods can show accuracy of 99% when used with a GPU in less than a second [4]. The region of interest (ROI) technique has also become very popular because of its simplicity. It is best suited for applications such as identifying whether a person is happy, sad, anxious, surprised, etc. [5]. Variations in facial landmarks such as eyes, pupils, eye lids, mouth and nose play a major role in identifying facial expressions. So localizing these facial land marks using ROI has given the best results in training the datasets. These methods usually depend upon CNN, support vector machines (SVMs), cascading algorithms, boosting classifier algorithms, etc. [6] and proposed an edge-based localization method using ROI selection.
User acceptance of virtual reality technology for practicing digital twin-based crisis management
Published in International Journal of Computer Integrated Manufacturing, 2021
Pak Ki Kwok, Mian Yan, Ting Qu, Henry Y. K. Lau
The prototype system for the practicing DT-based crisis management involves three core parts, namely, the graphical interface (mobile applications and virtual reality), computer simulation model, and database. Figure 2 shows an overview of the prototype system. In detail, the proposed system involves two server machines. One of them hosts the simulation model about the operation of a sophisticated system, whereas the other one hosts the database that stores various emergency scenarios and the trainees’ performance. The simulation model seamlessly obtains information from its physical counterpart such that trainees in the VR environment could be trained in a system that realistically behaves. Moreover, the prototype system also involves multiple personal computers, tablets, smartphones, and VR devices, which continuously display updated information about the emergency to the trainees. Trainees can also take action via those devices. Their commands are sent to the simulation model via the intranet, and the simulation model can instantly compute the development of the situation based on their actions.
Depth vision guided hand gesture recognition using electromyographic signals
Published in Advanced Robotics, 2020
Hang Su, Salih Ertug Ovur, Xuanyi Zhou, Wen Qi, Giancarlo Ferrigno, Elena De Momi
With the enormous growth of novel techniques and devices for human-robot interaction (HRI) [1,2], hand gesture recognition attracts increasing research interests for its provided benefits, such as advanced intuitiveness and ease use, for human-computer interaction (HCI). In particular, hand gesture recognition is widely applied in multiple applications, from virtual reality, computer games, health care, robot manipulation [3–5]. Especially for robot-assisted surgery, surgical robots have multi-degree-of-freedom due to the complexity and accuracy of medical operations, which requires intensive training and sophisticated manipulability for surgeons [6,7]. Therefore, hand gesture and electromyography (EMG) signals can be utilized to improve the interaction between human and surgery robots to be more intuitive [8–10].
Deep neural network approach for annual luminance simulations
Published in Journal of Building Performance Simulation, 2020
Yue Liu, Alex Colburn, Mehlika Inanici
While only a few previous studies existed in daylighting research, machine learning has been increasingly employed in closely related problems, such as rendering and appearance synthesis, in computer vision and graphics. Rendering a scene under novel lighting conditions has been a long-lasting research question in computer vision and graphics fields with applications in virtual reality, augmented reality, and visual effects. Machine learning has been investigated for image de-noising (Bako et al. 2017; Chaitanya et al. 2017). By creating high-quality images from noisy renderings with a reduced sample rate, the method has great potentials in accelerating physically-based rendering processes.