Recently, the International Virtual Reality Innovation Conference was officially opened in Qingdao Laoshan District Convention and Exhibition Center. The Virtual Reality Key Technology Trend Forum was held on November 10th. VR Network (WeChat Public Number: VR Platform) attended the site as a specially invited media.
The forum invited Chen Tong, deputy director of the ICT Manufacturing Research Department of the Institute of Information and Communications, China Institute of Information and Communications, to analyze some of the key issues of virtual reality in the early stage. The following are the contents of the speech:
Everyone in the virtual reality understands what this concept is and is an old new concept. At the same time, Huawei and its partners have different gradients in the quality of video transmission over the network. Our product classification has its own division. VR and AR are all integrated and host type. According to the recognition characteristics of human eyes, there are some non-light fields. Currently, the origins are mainly based on investment, hardware and software, and policy support. The relevant VR and AR are also in the park. There are some expansions. Under this circumstance, we may support ministries and commissions to briefly sort out the links, near-eye display, perceptual interaction, network transmission, and so on. This has a three-level list on the top five technology trends, such as near-eye display in key technology maps and among key technologies. There may be optical waveguides, screen renderings, etc. There are also key technologies such as network transmission.
There are some technical trends in these technical trends. For each technological trend, we will have a topic analysis. Why is it a technological trend? At present, there are several problems in the unification of technological trends. One is that the industry currently faces the VR field. The summary of trends in a certain technology field is too single, and there is development behind it. Why this trend is so important, the reason may be rather simple. Maybe every manufacturer or some industry people will feel that they will be very important. In fact, there will be One weight. For example, Chang’s classification of division is of course important. The sharpness of the right side and the black and white distance of the grid will make the border and screen window effect very small. Part of it is spacing, part of it is resolution, there are more grids, and black boxes are still there. Therefore, there are some dizzy controls on the AR. There are several reasons for this. On the one hand, we are familiar with motion sickness. Now there are several ways in the industry, such as taking medicine to solve this problem, and 5×. In the space of 5, the visual and physical consistency. There is also the picture quality that causes the 20 ms delay of the picture. There is trailing and flickering. Now the maximum time is still displayed, so there is no network transmission.
The third time delay is relatively small in the industry, and some guests also mentioned that it is to adjust the conflict, which is currently the industry to do more work. It was also mentioned that the emerging OLED field is brightness and contrast. There are also some interactive interactions that I will not introduce to you. However, based on the perception of interaction, I feel that if I want to propose a tracking from the inside to the outside, I think the most important thing The point of rendering, the reason why this is important, is often because the content of the screen is designed through the interaction of eye movements and content. In fact, this may not be the focus of a problem. Although the listening is very mysterious, the most important factor is the three aspects. Part of it is indeed the cost of the GPU, because this is the threshold and bottleneck. More importantly, there is a third dizzy problem in the near-eye display. It is the convergence adjustment conflict. In fact, the depth of field is changed. Therefore, there will be some problems. When people look at the outside world, they may focus on flowers when the outside perspective will be In fact, the physiological mechanism of this human eye is not implemented in the current near-eye display. However, if the fixation point of view is used, it may be able to do this, that is, based on the human eye to make an actual adjustment on the display side. , while reducing some of the vertigo, while we determine the point of view rendering and corners, plus the three areas of eye tracking is the rendering of the rendering process, as well as the eye tracking in the perceptual interaction and the zoom system in the near eye display. Together can solve the problem of dizziness.
In the field of cognitive interaction development, it is also focused on some machines to feel that problems, such as some biased semantic understanding, are the current problems of artificial intelligence and the problems of biased spatial geometry. For example, they did a job that is equivalent to the GPU and The GPS inside did an indoor VPS. The technology related to network transmission has also been mentioned by the contractor and will not be launched.
It may be added that the rendering of the gaze point just mentioned is really in line with the physiological characteristics of the human eye because the cells in the human eye are among the most central fundus in the human eye cells. There are two kinds of cells, and the pursuit of chasing cells is covered by the human eye. The high density is mainly responsible for the resolution and sharpness. The injecting cells are light sensitive to human eyes, which can reduce the workload of the GPU. On the load of the GPU rendering, some artificial intelligence methods and deep learning methods are used. Pixels that have already been rendered are based on this scheme to push out the surrounding pixels through deep learning, and then reduce the GPU's workload.
So we talk about the two major trends in rendering processing. One is open source and the other is interception. Open source is optimization design. The interception is mainly through some algorithm optimization. Some colleagues mentioned ADW and other asynchronous implementations to reduce the algorithm work. . There is also OpenX, which is a set of things because 3D is a very important task in rendering processing. Microsoft has its own platform. Apple also has its own 3D work. But now, it is independent, future or present. This can also be done in mobile phones because of 3D rendering support and cross-platform features, from mobile to non-mobile, while the system also has a more precise control of the GPU, because now all Mobile phone manufacturers have said that mobile phones are multi-core. In fact, I think this is not so important for GPU rendering. The reason is that there is no real development and utilization of multi-core CPU utilization. Through the API, it can be controlled independently by different threads. The real use of CPU resources, because the processing of the rendering, the GPU's workload is sometimes inadequate, but the rendering of the primary CPU also bear a lot of work, so this point of view, the CPU's multi-threaded use is very important work.
This is the development time and road map of the key technologies we actually talk about in various fields. These are the near-eye display and network transmission. We may have looked at the industrial ecology, and different colors have different industrial entities. The important thing is that there is a fusion between VR and AR. Before I heard about what VR is, what AI is, and the problem is VR+AI, so I always think about what "+" means. This is the era of mobile Internet to artificial intelligence. The focus of evolution, in this focus, I think that VR is still a relatively large development. Whether it is Google, Apple, Facebook, etc., it is said that this is a big trend. At present, the trends in the industry are really many. All say their own trends, this trend may be a real trend, rather than a pseudo-trend. Under the influence of this trend, virtual reality is firstly related to how the mobile Internet relates to artificial intelligence and how it relates to the mobile Internet. In two respects, one is in the current situation, or is mainly in the content of carrying VR and AR through the mobile phone, so see the mobile phone has been in VR and AR's bearer platform has always been more than 80% of the stock market in 2016. In addition, it does have some impact on the architecture of VR's mobile Internet, especially on cloud servers. It may feel better for the traffic of cloud servers and the architecture of cloud servers. There is a huge impact and challenges. At the same time, there are some challenges to the low power consumption of key components. Even if foreign news media frequently hold this matter, after the application of AI, the charging cable with a mobile phone can also be in 20 minutes. Power down by 20%, so if there is no charge, how bad this power will behave.
On the other hand, the impact of VR and AR on AI, which actually does some work on perceptual interaction and rendering processing, artificial intelligence, deep learning, and the other is the impact of VR on AI. In the face of this problem, some companies go through virtual The actual environment generates data. Virtual reality will create an environment in which some training data is reduced by the accumulation of some training data and artificial labeling costs through some generations in the virtual environment.
Now a company has done a better job in this area and has also taken in a lot of money. What they are doing is virtual reality + cloud computing + artificial intelligence. They have done a great job of collecting data.
From the perspective of industrial development, apart from these key areas of near-eye display, perceived interactions, etc., we think that in the future by 2020 we may feel that we may be experiencing interaction and content production, because then we ran up and it may follow 5G after 2020. Listing, network transmission will be the wave of the next wave of industries.
So, this is OpenX mentioned by colleagues and leaders just now. Every development person doesn't need to pay attention to what he is in development engine. He doesn't need to pay attention to what it is to run. He can directly direct the difference between the above content developer and peripheral manufacturers. The runner is working.
Similarly, there are similar industry alliances and standardization organizations in the field of cognitive interaction. To visually understand machine vision, this may be the work of training and reasoning in deep learning. In fact, this is still the reasoning of cloud training for local reasoning. The problem of fragmentation, the same standardization organization may use this format to harmonize the training and reasoning formats.
At present, from the perspective of intellectual property rights, we have seen problems. We have also seen problems in terms of quantity and quality. There are many designs and patents, and there are even more patents in universities than in enterprises. In summary, there will be some work and performance in various fields. The gap between our country is small and large. It may be that the rendering process will be larger, the near eyes will be better in the optics, and the display is also quite good, but it does There are some advantages of Google's platform and the advantages of network transmission. Huawei and ZTE do better.
There are also AR and AR cases, such as concerts, the military, and several scenes of car networking. One is the way the car is connected through the chip, and there is also a way to display the headlights. How can the front traffic lights be seen? There is also a depot in the design and manufacturing integration, there will be some work. In the medical field, everyone may be familiar with it. Apart from some rehabilitation exercises and early detection, the industry will mainly mention more. Perhaps the larger trend in the future is the digital twin industrial Internet, because after all, the giants of industrial software have also received Qualcomm, so also made a separate part.
Therefore, the trend is still a big development trend, so it mainly consists of two parts, some of which are beyond, some of which are principles. We also hope that everyone can surpass what they want to go beyond VR and thank you!
This article is original by VR net, please indicate VR net and chain back.