Does the video have a boundary?
In the past, the answer was yes.
At that time, the video would be locked in the TV, locked on the big screen. But as more and more hardware devices enter the home, as the network technology changes, as the continuous upgrade of computing power, as the video codec ability continues to improve… Video has become a new information carrier, and as the base of the big video industry in the new era, video cloud has been given the key mission to change the society.
On July 10, “Imagine” — 2021 AliYun Video Cloud Panorama Innovation Summit and global Video Cloud Innovation Challenge final award ceremony was held in Beijing. About the future development prospect of video, about the panoramic blueprint of video cloud, about the interaction of academia, art and venture capital, about the multi-dimensional exploration of developers and audio and video technology, through the collision of views of this summit, we seem to have a glimpse of the future development of video.
From video to hypervideo, the video cloud role is added
In the last few years, the word video-ification has been used more and more. So what is video?
To put it simply, the transmission of information is gradually added to the video as a carrier. Due to the continuous reduction of the video production threshold, the continuous improvement of user acceptance, and the increasing use time of users, the era of full video content has arrived. At the same time, video not only shines in the consumer sector, in education, conferences, healthcare, finance and other industries are also generating new demand.
The overall time consumption of users spending on video has increased significantly, as has the overall social video-based interaction in a wide variety of business scenarios. “Content is evolving more towards video, interaction is more diverse, compared to the previous video, this is a hypervideo era.” Ali Baba researcher, Ali cloud intelligent video cloud head Lin Hao so the current definition of these changes.
To define an era, you need to understand it. Lin hao explained that the hypervideo era has five characteristics: hyper-content, hyper-interaction, hyper-link, hyper-language ability and hyper-future vision. From the perspective of analysis, it means that the form of video continues to evolve, the interaction is more rich, the spread of the language limits, and can also affect the daily life of the public through AR, VR and other ways.
So how did the era of hypervideo come about? Lin Hao thinks 5G has played an important role in promoting, 5G large bandwidth to promote AI and IoT development for intelligent networking; 5G activates uHD video and VR/AR, making the peak rate of the network reach 20Gbit/s, the wireless interface delay is 1ms and the resolution is significantly improved. More importantly, 5G has opened up new formats of digital content. Whether it is digital games, interactive entertainment, film and television animation, three-dimensional video or digital performance, the performance ability and form of video have been greatly enriched.
More importantly, cloud + video serves as a catalyst for scene innovation, making the combination of virtual and real possible. The integration of cloud edge and end makes the edge computing power move up and the cloud computing power sink down, which reduces the processing pressure and time delay. The cloud-integrated audio and video technology makes it possible to have a consistent dual-end experience; After the development of AI technology, the power of the whole video link is realized, and the intellectualization subverts the previous content production mode; At the same time, mixed reality technology has also broken through new forms of content and interaction, making the last barrier between the physical world and the digital world broken and linked, making video a carrier with more possibilities.
Steve Jobs once said, “On lower bandwidth, people will transmit messages. On higher bandwidth, people will transmit emotions.” The era of hypervideo is born not only because of bandwidth improvements, but also because of technological evolution.
The evolution of technology can be divided into two directions: the evolution of content and the evolution of interaction. The evolution path of content follows four characteristics of greater density, more dimensions, more senses and topological space, and its specific performance forms from text, image, video, live short video, information and knowledge, and even the whole scene content, and finally forms the immersive content form. The evolution of interaction follows the characteristics of multi-end links, multi-person sharing, space breaking and seamless integration of virtual and real, and the evolution path forms the process from offline, online and interactive full-scene online to immersive interaction.
It’s not hard to see that immersive interaction and content forms are the real future we can explore. “Information will naturally pass from one interaction object to another. The numbers will coexist with and enhance the physics.”
Interaction like Ready Player One is not a fantasy. Of course, behind all imagination is the deep mining of technology. Behind video is not the upgrade of AI, data, codecs and other single-point technologies, but the construction of the entire technology system based on video cloud. Video cloud is not only a cloud technology, but also a continuous evolution in the overall technology of video. Whether it is three-dimensional or holographic, it should keep evolving and layout, and finally make more combination of video and scene, so as to realize “cloud innovation and value creation” powered by digital audio and video.
Big video industry base, video cloud industry evolution
With the development of hypervideo, the Internet is also developing. The value of an industry is no longer measured by the number of devices, but by the number of hours. While dividends in all areas of the Internet have almost dried up, the video-related sector showed huge dividends last year. And xu Fanlei, deputy general manager of iResearch, said the dividend will continue.
From the perspective of industry development, the present stage of the video industry is fragmented, decentralization, Gao Qinghua, real-time and so on a series of characteristics, that is in demand side for the pursuit of video became more “short, frequency, fast”, the pursuit of perfection quality experience, the need of real-time audio and video, real-time interactive remodeling video application value, Thus covering financial services, healthcare, public utilities, social networking, education, consulting and many other industries.
But if we look at the process of information transmission in human history, video plays an important role. Initially, human communication was more about body language, which was physically demanding and ambiguous. Then we had language, which had no physical problems but was bound by space and time and was difficult to pass on; Later, we had text, and it was easy to inherit for a thousand years. However, the inherent threshold of text and the lack of information richness prompted the emergence of video. And video continues to evolve, from TV, to offline player video, to live audio and interactive video.
However, video is still not perfect. There are two main problems with video. First, it is linear. The second is that the modification is slower and more difficult than the text. And based on these problems, the industry will be more and more integrated with video. That is to say, video is no longer an industry, but a kind of underlying basic capability, based on the video cloud to build video applications will become a mandatory option. As video becomes a must, it can be said that “video cloud is the base of the big video industry in the new era”.
The impact of the deep integration of industry and video goes beyond products, changing the landscape of many industries. However, due to the complexity of the industry, their demands for video capabilities are different but have some similarities. The first is that it is easy to integrate, easy to measure, it needs to be less costly, more flexible to scale to try on the cloud, and it needs to be able to go into production quickly with agile trial and error.
Therefore, the video cloud needs to provide different solutions and process support in different links such as production, processing, transmission and consumption. In addition to the depth and division of the video itself, the cloud can significantly reduce the barriers to producing high-quality, valuable video.
In this process, cloud services are extremely important for supporting video. In the video production process, the video cloud can provide intelligent content processing capability, greatly improve the creation efficiency and realize efficient media asset management. At the processing stage, the video cloud achieves the optimal balance between cost and quality through video processing and intelligent coding. In the transmission link, the video cloud is intelligently accelerated based on CDN, and the cloud side and end cooperate to reduce transmission delay and save bandwidth cost. In the final consumption link, video cloud can also provide beauty, bel canto, immersive interaction and other diversified gameplay to enrich user experience.
The video cloud itself continues to evolve in combination with the industry. At present, although the video cloud is mainly concentrated in the Internet and pan-entertainment field, it already has the ability to provide support in different links, and can continue to evolve in various industries. At the same time, video cloud solutions also give users more choices, whether application-level capabilities, or industry-wide generic platform enterprises, different dimensions and different users can have different answers.
In addition, video cloud is still pursuing perfection in technology. Although it is not mature enough to solve the problems of HD, real-time and interactivity, the concept of software defining everything is cooperating with hardware to deal with many links such as router, storage and computing. At the same time, low-code development also appears in a large number of video cloud and video industry, which can enable practitioners to call functions more quickly and agile, improve the ease of use, and realize easy to call and easy to integrate.
In the future, video-based cloud is likely to create more innovations, which can provide users with more links, lower barriers to entry, and show more universal energy. The video cloud technology for the whole video industry and large video industry, is to become a base function.
Sustainable development of video cloud, technical difficulties and breakthroughs
Video cloud as an industrial base, one of the major characteristics is compatible and package. Especially at present, users’ demand for video interactivity, presentation mode and immersive experience is increasing. The deep integration of AI will become the key to the innovation of video cloud and video industry. While the video cloud is expanding in social, entertainment, education and other fields, deep learning continues to play a great role in image, voice, language, big data feature extraction and many other aspects. It can be said that future breakthroughs in video cloud technology will be driven to some extent by artificial intelligence based on deep learning.
Round table at the end of the activities in the BBS, intelligent information processing laboratory, institute of computing technology, Chinese Academy of Sciences, said researcher Wang Shuhui, deep learning age brought the rise of artificial intelligence for the third time, the rise mainly used for the purpose of making deep learning technology has a good effect in many tasks, but its kernel problems. Therefore, in order to achieve a breakthrough in video technology, three technical problems should be solved from the perspective of the internal mechanism of deep learning.
- First, the existing deep learning relies too much on data, and its data processing performance and knowledge utilization are insufficient. Therefore, based on this consideration, knowledge building of network multi-modal cross-media data will be an important development direction in the future.
- The second is to build a knowledge base to support the reasoning of machine systems, so that the machine can draw inferences from any number of different sources.
- Third, in the early days, people were not equal to computers, such as human-computer collaboration in content creation. In the core process, algorithms, systems and people need to be trusted, and mutual trust, collaboration and reliable reasoning will be the main problems to be solved.
Of course, AI, while problematic, also plays an important role in video. Xie Xuansong, a senior algorithm expert at the Dharma Institute, said that the role AI plays in video is mainly divided into two categories. The first category is the most basic video or image understanding, including classification, marking, detection, segmentation and so on. The second category is related to the production category, such as production, editing, processing, erasure, erasure, etc., which also includes the underlying visual related enhancement, etc.
Image enhancement of video is a major application direction of AI. When the resolution is low, the information experience of video will be very poor, and more vivid colors will also increase the experience. More immersive experiences are the way to go. Detail, smoothness and color, for example, are important things to focus on if you want to create 4K content. But from the technical point of view, the following three problems must be directly faced. First, the more the pursuit of details, the more likely there will be defects. How to ensure the reduction of details and the control of defects, which is a very core technology; Second, the source of the algorithm is data. There are generally two sources of data, such as low resolution and high resolution, low picture quality and high picture quality. Finally, data acquisition often needs to be solved by manual method with high cost, which is also a big difficulty. Third, in the practice of AI technology, the balance between effectiveness and efficiency is also a problem.
AI is also moving in two dimensions. One is to serve consumers, and the other is to reach into all industries to reduce costs and improve efficiency, and create all kinds of opportunities.
Ultimately, of course, it is still people who drive innovation and technology upgrading. So the popularity of AI has been high for many years, and many schools have started AI related talent and education, but for the market and industry, the problem of talent shortage is still serious, so where do the talent go? Wang Shuhui said that most of his graduate students have joined the battlefield of the industry, and the university has sent a large number of talents to the industry. However, because the industry is developing too fast, high-level talents are scarce, and different laboratories have different positioning, so they cannot blindly expand their scale.
At the same time, laboratory research is to separate problems from reality and solve them through mathematical methods. However, enterprises have different requirements for students. They want enterprises to understand the business and put it into practice. There is a long chain from academic research to business applications, which makes it difficult for students to plug and play. And it’s clearly not just schools that are aware of this, but industries and businesses as well.
This year, Ali Cloud jointly hosted by Intel, and youku strategic technology cooperation global video Cloud Innovation Challenge held the final award ceremony at the summit. The contest by tianchi platform and ali cloud video cloud, focusing on video cloud technology in the field of application and innovation of the industry, attracted 23 countries around the world, more than 4000 teams participating team, competition is divided into “algorithm” and “innovation” two series, fully explore talents, encourage and look forward to the contestants inspire more imagination in the future.
In addition, ali YunTianChi platform at the summit also released tianchi data sets, open source project, covers electricity, finance, logistics, healthcare, energy and so on more than 60 have real business scenarios industry scarce data sets, hope that through open real business scenarios and data, and all the social forces to create a professional scientific research data platform.
The development of video cloud has become the choice of The Times, but also changes the business and society, into a large video industry base; Video cloud technology can be full of imagination, break through time and space, and make the communication between people more seamless and comfortable.
The future has arrived, a new video world, are you ready?
All the speech contents of this video Cloud Panorama Innovation Summit will be released in the “video Cloud Technology” public account successively.
“Video cloud Technology” is your most noteworthy audio and video technology public account. It pushes practical technical articles from ali Cloud every week, and communicates with first-class engineers in the field of audio and video. 【 Technology 】 You can join ali Cloud video cloud product technology exchange group, and the industry to discuss audio and video technology, to get more industry latest information.