Editorial issue 32.2

Abstract

This issue contains six papers. In the first paper, Dong Yang, Wenjing Yang, Minglong Li, and Qiong Yang, National University of Defense Technology, in Changsha, China present a role-based attention model for reinforcement learning. The proposed model uses convolutional neural networks to generate soft attention maps, adding crucial role information in the task, forcing the agent to focus on important features, and distinguish task-related information. To validate the performance in complex problems, the proposed approach is evaluated in a challenging scenario, Football Academy in Google Research Football Environment, a newly released reinforcement learning environment with physics-based 3D simulator. The experimental results demonstrate that agents using role-based attention mechanism can perform better in football games. In the second paper, Jie Liao, Yanping Fu, and Chunxia Xiao, from Wuhan University in China, Mengqiang Wei, from Guangzhou Boguan Telecommunication Technology Limited, in Guangzhou, China, and Qingan Yan, JD.com Inc, in Mountain View, USA, propose a new approach for performing high-quality dense Multi-View Stereo. Although many Multi-View Stereo (MVS) algorithms have been proposed, constructing a dense 3D representation in textureless image region is still a challenging problem. The key insight of their solution is based on the image texture enhancement for the unstructured input images. For each image, they first perform the multiscale bilateral decomposition and reconstruct the multiscale texture-enhanced image. Then, they perform Multi-View Stereo on both the enhanced images and the original input images. Subsequently, they merge the corresponding output depth and normal maps. Finally, they propose a novel selective joint bilateral propagation filter to produce more dense 3D models. In the third paper, Alessandro Bruno, Morgan Moore, Jinglu Zhang, Stéphane Lancette, and Jian Chang, from Bournemouth University, in UK and Ville P Ward, from Shoppar Ltd, in London, UK propose a novel technique based on Head Movement tracking to explore multi-layer digital content. They extend an existing method by Kazemi et al. dealing with the extraction of facial landmarks to define the ‘head-gaze’ of the user. They use the ‘head-gaze’ to calculate the users’ on-screen coordinates. Hovering the cursor over an interactive area for a given time threshold allows users to explore the next layer contents. Their experimental sessions allowed them to measure the technique’s level of control and usability. Their results were promising, and users were able to interact with considerably small regions. Furthermore, their lightweight method can be used with a low-cost camera or webcam and a wide range of screen sizes and distances. The fourth paper, by Andrzej Grabowski and Kamil Jach, from Central Institute for Labour Protection, in Warsaw, Poland, presents a brief history of the development of virtual reality (VR), followed by a comprehensive literature overview on the matter as well as an example of the application of VR in the training of professionals on the example of firefighters. The literature review presents various examples of the use of VR in the training of professionals coming from multiple fields, such as medicine, psychology, forestry, iron casting, or construction. The case study shows it can be used to recreate even the most complex scenarios and surroundings, even those of the firefight in confined spaces where enclosed fires tend to appear and have been considered vastly hazardous for not only civilians or equipment but also to firefighters dispatched to contain them. In the fifth paper, Benjamin Stephanus Botha, Lizette de Wet, and Yvonne Botma, from the University of the Free State, in South Africa present a study in Virtual Reality for education. During this study, the researchers created a virtual environment for use in Southern Africa, where students could practice managing a young adult with a foreign object in the airway. The aim of the virtual environment was to determine whether a viable, “home-made” solution could be created, which could also be expanded later on to incorporate more scenarios. This was due to the expensive nature of existing systems for virtual clinical simulation. To determine whether the virtual environment is usable, two expert review panels assisted in testing the virtual environment. The first-panel being Computer Science experts and the second Health Science experts. Each panel evaluated the environment and the scenario using heuristic evaluation and cognitive walkthroughs. The findings and recommendations made during the expert reviews are presented in this paper to assist in improving future developments within the field of virtual reality in Health Science education, especially for developing countries in Africa.