A research article titled “Personalized 360-Degree Video Streaming: A Meta-Learning Approach” by the University of Michigan-Shanghai Jiao Tong University Joint Institute (UM-SJTU JI, JI hereafter) Assistant Professor Yifei Zhu and undergraduate student Yiyun Lu has been accepted by the 30th ACM International Conference on Multimedia (ACM Multimedia hereafter), a worldwide premier conference and a key world event to display scientific achievements and innovative industrial products in the multimedia field.
Yifei Zhu is the corresponding author and Yiyun Lu is the first author of the paper. This paper presents the first meta-learning-based personalized 360-degree video streaming framework. Through efficient meta-network designs, the commonality among viewers of different viewing patterns and Quality of Experience (QoE) preferences is captured to optimize viewer’s QoE.
360-degree videos have attracted wide interest for the immersive experience they bring to viewers. The rising of high-resolution 360-degree videos greatly challenge the traditional video streaming systems in limited network environments. Given the limited bandwidth, tile-based video streaming with adaptive bitrate selection has been widely studied to improve the QoE of viewers by tiling the video frames and allocating different bitrates for tiles inside and outside viewers’ viewports. Existing solutions for viewport prediction and bitrate selection train general models without catering to the intrinsic need for personalization.
The accepted paper presents the first meta-learning-based personalized 360-degree video streaming framework. The commonality among viewers of different viewing patterns and QoE preferences is captured by efficient meta-network designs. Specifically, the research team designs a meta-based long-short term memory model for viewport prediction and a meta-based reinforcement learning model for bitrate selection. Extensive experiments on real-world datasets demonstrate that their framework not only outperforms the state-of-the-art data-driven approaches in prediction accuracy by 11% on average and improves QoE by 27% on average, but also quickly adapts to users with new preferences with on average 67%-88% less training epochs.
ACM Multimedia is an influential academic conference in the field of multimedia processing, analyzing and computing sponsored by the Association for Computing Machinery (ACM). It is also an A-level conference recognized by China Computer Federation (CCF). Since its inception in 1993, the event is held once every year. The 30th ACM Multimedia to be held in the Portuguese capital Lisbon from October 10 to October 14 received submission of 2473 papers. A total of 690 papers were accepted after the reviewing process.
Background Information
Yiyun Lu is a senior student of JI majoring in Electrical and Computer Engineering. He is expected to pursue his Master in Computer Science degree at Brown University in the United States upon graduation this year.
Yifei Zhu is currently an assistant professor of the University of Michigan-Shanghai Jiao Tong University Joint Institute. He received the B.E. degree from Xi’an Jiaotong University, Xian, China, in 2012, and the M.Phil. degree from The Hong Kong University of Science and Technology, Hong Kong, China, in 2015, and the Ph.D. degree in Computer Science from the Simon Fraser University, BC, Canada, in 2020. His current research interests include edge computing, multimedia networking, and distributed machine learning systems. This work is supported by SJTU Explore-X grant.
