Efficient Scheduling for Dynamic Streaming of 3D Scene for Mobile Devices
Download as PPTX, PDF0 likes71 views
The document discusses an efficient scheduling approach for dynamic streaming of large 3D scenes optimized for mobile devices, utilizing cloud-based geometry loading. It highlights a scalable client-server architecture that facilitates the streaming of geometries based on user-selected paths and bounding box processing for improved pathfinding. The study also presents experimental results demonstrating the advantages of scheduling in reducing loading times and rendering delays while suggesting future enhancements for camera dynamics and texture optimization.
![Related Work
• Precomputation approach
– Precompute geometry importance : Tian and AlREgib [1]
– Precompute illumination : Pacanowski et al. [2]
• Regular scene subdivision Wang et al. [3]
• Optimize texture streaming : Eu et al. [4] and Englert et al. [5]
• Network approach
– Attach geometry to other file format : Concolato et al. [6]
– Distributed P2P streaming: Wang et al. [7] and Jie et al. [8]](https://image.slidesharecdn.com/iwait2018presentationntutheme-novideo-190221062244/85/Efficient-Scheduling-for-Dynamic-Streaming-of-3D-Scene-for-Mobile-Devices-5-320.jpg)
![Future Work
• Consider varying camera speed and orientation
• Consider perceptual factor
• Progressively stream vertices based on mesh refinement method
[2].
• Optimize texture by using perceptual-guided refinement strategy [4]
• Upscale the test data and using more robust path-planning method,
e.g. subregion graph [13]
• Test on standalone VR devices
– HTC Vive Focus
– Oculus Quest](https://image.slidesharecdn.com/iwait2018presentationntutheme-novideo-190221062244/85/Efficient-Scheduling-for-Dynamic-Streaming-of-3D-Scene-for-Mobile-Devices-18-320.jpg)
Efficient Scheduling for Dynamic Streaming of 3D Scene for Mobile Devices
- 1. Efficient Scheduling for Dynamic Streaming of 3D Scene for Mobile Devices Budianto Tandianus1 , Hock Soon Seah1 Tuan Dat Vu2, Anh Tu Phan2 1. Multi-plAtform Game Innovation Centre, Nanyang Technological University, Singapore 2. School of Information and Communication Technology, Hanoi University of Science and Technology, Vietnam
- 2. Introduction • Visualization of very large scene using out-of-core approach – Client application does not have enough resource to load all data • Aim for scalability – Support for mobile devices • Cloud-based geometry loading approach – Geometry is stored in cloud and it is streamed to client based on proximity to camera – However, objects with large file size may delay streaming of other objects
- 3. Introduction • Optimize streaming by analyzing the user-selected path • Basic client-server architecture – Client (mobile devices) sends starting and finishing points to server – Server (storage and compute units) returns path and geometries along the path – The geometries are streamed according to the determined schedule • Client receives the geometry and import it to the scene – Client implemented in Unity – Geometry in FBX format
- 4. Contribution • Scalable and efficient scheduling scheme for cloud-based urban simulation approach – Support very large scene in mobile devices – Can be used in various applications such as visualization and games – Designed to handle non-urban objects (e.g. vehicles and humans), not only urban objects (e.g. buildings) • Realistic mobile walkthrough applications – Car navigation – Historical site walkthrough – Large complex (e.g. university) navigation application for visitors
- 5. Related Work • Precomputation approach – Precompute geometry importance : Tian and AlREgib [1] – Precompute illumination : Pacanowski et al. [2] • Regular scene subdivision Wang et al. [3] • Optimize texture streaming : Eu et al. [4] and Englert et al. [5] • Network approach – Attach geometry to other file format : Concolato et al. [6] – Distributed P2P streaming: Wang et al. [7] and Jie et al. [8]
- 6. Proposed Method • Data preprocessing : – Each object (e.g. building) is a separate FBX file and it is processed into an asset bundle – Also obtain position and bounding box of each object – The bounding boxes are grouped • Path finding : Djikstra’s algorithm – The grouped bounding boxes are treated as obstacles
- 7. Bounding Box Processing Extracted bounding boxes. Grouped bounding boxes.
- 8. Path Finding Path 1. Path 2.
- 9. Streaming scheduling • Sample points at regular interval in the path • Define a rectangular loading area centered at each point • 3D objects fall into the loading area are organized into a set and they are sorted from largest to smallest in term of file size • The sets are combined sequentially
- 10. System Design • Cloud storage : Internally developed cloud storage • Mobile client : developed by using Unity • PHP web server : Yii2 • Python service : Process bounding box by using Shapely. Also, use pyvisgraph to create visibility graph for path finding • RabbitMQ : message broker
- 11. System Design
- 12. Experiment Setup • NTU dataset – 9,629 bundles – Total 932 MB • Mobile device specification: – Samsung Galaxy S7 Edge with Exynos 8890 CPU and 4GB RAM • Compare with and without scheduling – With scheduling: use RabbitMQ and scheduling algorithm – Without scheduling: do not use RabbitMQ and scheduling algorithm. Assets are transferred directly from server to mobile client. Client always requests to server objects within vicinity.
- 13. Result
- 14. Result
- 15. Result With scheduling. Without scheduling.
- 17. Conclusion • Scalable • Mitigate late-loading of 3D models • Rendering time advantage • Scheduling method will consume larger memory only during earlier traversal
- 18. Future Work • Consider varying camera speed and orientation • Consider perceptual factor • Progressively stream vertices based on mesh refinement method [2]. • Optimize texture by using perceptual-guided refinement strategy [4] • Upscale the test data and using more robust path-planning method, e.g. subregion graph [13] • Test on standalone VR devices – HTC Vive Focus – Oculus Quest
- 19. Acknowledgement • This research is supported by the National Research Foundation, Prime Minister’s Office, Singapore under its IDM Futures Funding Initiative.
- 20. The End