Dynamic Splat Streaming for Viewport-Based Bandwidth Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Three-dimensional (3D) models require significant computational resources and bandwidth due to unnecessary fidelity in rarely viewed parts, leading to streaming and rendering overhead and delays.
Innovation Solution
Dynamic streaming of splats based on user viewing patterns, where parts frequently viewed are rendered at higher fidelity with more splats, and less important parts are rendered at lower fidelity with fewer splats, optimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high fidelity splat representations are used for all parts of the 3D model, then the quality and detail of the model visualization is improved, but the bandwidth consumption and streaming overhead increase significantly
Solution Approach 1:
The patent applies local quality by assigning different fidelity levels to different regions of the 3D model based on their importance. High-fidelity splat representations are applied only to important regions that require detailed visualization, while low-fidelity representations are used for less important regions. This resolves the contradiction by maintaining high visualization quality where needed while reducing overall bandwidth consumption through selective application of fidelity levels.
Solution Approach 2:
The patent segments the 3D model into multiple regions with different importance levels. By dividing the model into important and less important regions, the system can apply different splat representation strategies to each segment. This segmentation allows the system to optimize bandwidth usage by transmitting high-detail data only for critical segments while using compressed or lower-detail representations for other segments.
2Manufacturing precision
If high fidelity splat representations are used for all parts of the 3D model, then the detail and accuracy of the model is improved, but the rendering resources and compute power required increase
Solution Approach 1:
The patent reduces rendering resource consumption by applying high-quality splat representations only to important regions of the 3D model rather than uniformly across the entire model. The rendering system processes high-fidelity data selectively for regions that contribute most to the user experience, while using simplified rendering paths for less important regions. This local quality approach maintains model detail accuracy where it matters most while significantly reducing overall rendering compute power requirements.
3Stability of the object's composition
If uniform fidelity is applied to all parts of the 3D model, then the consistency of quality across the model is maintained, but the streaming efficiency and resource utilization decrease due to unnecessary data transmission
Solution Approach 1:
The patent implements dynamic fidelity adjustment by determining importance levels for different regions of the 3D model and adapting the splat representation quality accordingly. The system dynamically selects between high-fidelity and low-fidelity representations based on region importance, allowing quality consistency within each importance category while optimizing streaming efficiency through selective data transmission. This dynamic approach resolves the contradiction by maintaining appropriate quality consistency rather than uniform quality across all regions.
Data Source
AI summary
A system and associated methods select splats at different fidelities to optimize a requested field-of-view of a three-dimensional (3D) model for presentation on a local device or for streaming to a remote device. The system determines priority values that are associated with parts of the 3D model within the field-of-view, selects first splats that represent a first part of the 3D model at a first fidelity in response to a first priority value being associated with the first part, selects second splats that represent a second part of the 3D model at a second fidelity in response to a second priority value being associated with the second part, and presents or streams the first splats and the second splats in response to the request in order to generate the field-of-view with different fidelities for the first part and the second part of the 3D model.