PARIMA: Viewport-adaptive 360-degree Video Streaming
With increasing advancements in technologies for capturing 360° videos, advances in streaming such videos have become a popular research topic. However, streaming 360° videos require high bandwidth, thus escalating the need for developing optimized streaming algorithms. Researchers have proposed various methods to tackle the problem, considering the network bandwidth or attempt to predict future viewports in advance. However, most of the existing works either (1) do not consider video contents to predict user viewport, or (2) do not adapt to user preferences dynamically, or (3) require a lot of training data for new videos, thus making them potentially unfit for video streaming purposes. We developed PARIMA, a fast and efficient online viewport prediction model that uses past viewports of users along with the trajectories of prime objects as a representative of video content to predict future viewports. PARIMA uses a pyramid model for Bitrate allocation, while providing the highest bitrate at the viewport.
Key Features
- Network-adaptive: The proposed approach aims to reduce the network load during 360° video streaming by considering a protocol that can reduce the amount of data to be transferred over the network based on the current bandwidth scenario.
- Viewport-adaptive: The method dynamically predicts the user viewport and provides the highest quality streaming around the viewport, improving the user perceived quality of experience(QoE).
- Lightweight ML model: Using PARIMA, we have achieved an average QoE improvement of around 35% and 78% over two baselines and an average improvement of 117% in adaptivity over a non-adaptive bitrate allocation scheme. Our model is lightweight and exhibits a prediction latency of under 1 second for a chunk size of the same duration.
- Large-scale testing: We evaluate our model on two publicly available data sets, one consisting of 5 videos with head movement data for 59 users, while the other consisting of 9 videos watched by 48 users, each video having a wide range of static and moving objects. We have made our code public for the research community.
Licensing:The platform source and the collected dataset is free to download and can be used with MIT License for non-commercial purposes.
Contributors
Sarthak Chakraborty
IIT Kharagpur, India
Lovish Chopra
IIT Kharagpur, India
Abhijit Mondal
IIT Kharagpur, India
Sandip Chakraborty
IIT Kharagpur, India
Publications
- Lovish Chopra, Sarthak Chakraborty, Abhijit Mondal, Sandip Chakraborty:"PARIMA: Viewport Adaptive 360-Degree Video Streaming", The Web Conference (Erstwhile WWW) 2021