In the field of audio and video applications, we have accumulated a large amount of user-reported data, including audio and video quality data, user behavior data, etc. These data are not just numbers, they are the link between us and our users. By deeply analyzing this data, we gain insights into our users’ needs, experiences, and behaviors, providing us with valuable insights.
By analyzing the audio and video quality data, we can understand the problems encountered by users during use, such as blurred pictures, sound delays, etc. This kind of data can reveal problems such as network stability, server load and device performance, help us accurately locate and solve various technical problems, and improve the stability and clarity of audio and video transmission.
Secondly, user behavior data tells us about users’ habits, interests and preferences in the application. By deeply understanding user behavior, we can customize recommended content for users, provide video recommendations that are more in line with user tastes, and increase user stickiness and satisfaction.
In this topic, we will listen to how cloud music video image technology and HDR Vivid improve content quality and enrich user experience; and learn about visual quality evaluation models and underwater image quality enhancement. Through data-driven, we can carry out targeted optimization to provide a perfect experience that is closer to user needs.
01
Cloud music video image technology application
Cai Miaomiao
NetEase Cloud Music Computer Vision Algorithm Expert
With the rapid development of the Internet and the widespread popularity of smart devices, the demand and consumption of video and image content has shown an explosive growth trend. This phenomenon is particularly prominent in the field of cloud music. Based on its huge user group and traffic base, cloud music continues to derive diversified video and image technology needs. These technologies not only play a key role in meeting increasingly updated content ideas, but also provide strong support for ever-changing social interaction gameplay. At the same time, these technologies also play a pivotal role in the efficient processing of huge background data.
This sharing will provide an in-depth discussion of which video image technologies are used in Cloud Music and how these technologies play a vital role in various businesses of Cloud Music. We will introduce in detail the specific applications of these technologies in improving user experience, enhancing music visualization, optimizing social interaction, and efficiently processing data.
02
One against ten - construction of efficient universal visual quality evaluation model
Wu Qingbo
Associate Professor, University of Electronic Science and Technology of China
With the rapid development of image and video business forms and product functions, the requirements for visual quality evaluation tasks are constantly changing. The traditional expert system-based evaluation model has a small scope of application, high storage costs, and high overhead for repeated construction, which creates a huge contradiction with the development needs of lightweight and fast iteration of image and video application products.
This sharing starts from two aspects: model structure and data annotation, and explores solutions for building efficient and universal visual quality evaluation models. The first part introduces the dynamic scalable network structure and how to use the continuous forgetting and learning mechanism to achieve one-to-ten evaluation model, that is, the same model serves the quality evaluation requirements of different tasks. The second part introduces the subjective bias inference of data annotation, and how to use the alternating optimization strategy to achieve one-for-ten data annotation, that is, only one person can annotate the same image to train a reliable quality evaluation model. Through the sharing and introduction of the above two parts of the solution, we hope to provide reference and inspiration for the development of low-cost, high-efficiency image and video application services.
03
The application practice of HDR Vivid in Tencent Video Zhencai Audiovisual
Feng Zhiping
Senior Development Engineer, Tencent Video Playback Technology Center
Currently, video platforms regard high-quality content as an important way to improve users' high-quality experience and user membership. With the development of HDR technology, HDR technology has been widely used in the field of high-quality content. Among the many HDR technologies, HDR Vivid, as a domestically constructed standard, has received end-to-end support from chips, terminal equipment, content production and production, platforms, codec systems, etc. As the executive director unit and content platform of the UWA Alliance, Tencent Video has participated in the construction, promotion and application of the HDR Vivid standard.
This sharing will be divided into three parts. The first part introduces the many current HDR technical standards, the difference between the HDRVivid standard and other standards, and the current ecological status of HDRVivid. The second part introduces Tencent Video’s layout in high-quality content and why it chooses HDRVivid as an important part of high-quality content. The third part introduces Tencent Video’s specific practices in HDRVivid content production and terminal consumption, as well as terminal certification.
04
Research on underwater image quality evaluation and image quality enhancement
Zhao Tiesong
Professor and doctoral supervisor at Fuzhou University
The ocean occupies more than 70% of the earth's area. However, there are still many problems in human exploration of the ocean. For example, optical imaging has many problems such as blur and color distortion, and the clarity of the picture also has a decisive impact on downstream computer vision tasks. There are still big challenges in how to effectively enhance the quality of underwater images and evaluate the image quality after enhancement.
This sharing mainly includes some of our research work on underwater image quality evaluation and image quality enhancement in recent years. The first part mainly introduces the research work on underwater image quality evaluation, including its implementation in underwater optical imaging and sonar imaging; the second part mainly introduces the research work on underwater optical image enhancement and repair, especially how to effectively combine AI technology , to improve the system performance of related tasks; the third part mainly introduces the algorithm integration of the underwater optical image processing system, including the above-mentioned quality evaluation, enhancement and encoding and decoding algorithms. Through the above three parts, the application of current image quality evaluation and enhancement algorithms in underwater imaging systems is systematically introduced.
LiveVideoStackCon 2023 Audio and Video Technology Conference Shenzhen Station
You are cordially invited to participate!
Time: November 24-25, 2023
Location: Shenzhen Sentosa Hotel (Jade Branch)
Inquiry: 13520771810 (same number on WeChat), [email protected]
