3D Model Streaming With Gaussian Splats for Uneven Resolution Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data networks have limited bandwidth and latency, making real-time or low-latency transfer of 3D models challenging, especially when multiple models are streamed in dynamic environments, leading to uniform degradation of visual quality across the entire model.
Innovation Solution
An adaptive streaming system uses a tree-based representation of 3D models with Gaussian splats, allowing non-uniform resolution adjustments based on network and rendering performance to preserve quality in important regions while reducing data in less critical areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If 3D model is streamed at high resolution, then visual quality is improved, but data transfer bandwidth requirement increases
Solution Approach 1:
The patent divides the 3D model into multiple resolution levels (e.g., 1080p, 720p, 480p) and organizes them in a hierarchical structure. This allows the system to segment the model data and transmit only the necessary resolution level based on network conditions, resolving the contradiction between maintaining high visual quality and reducing data transfer volume.
Solution Approach 2:
The patent applies different resolution levels to different regions or parts of the 3D model based on their importance and visual impact. Critical regions maintain high resolution while less important regions use lower resolution, thereby optimizing the balance between visual quality and data transfer requirements.
2Manufacturing precision
If 3D model is streamed at high resolution, then visual quality is improved, but network latency increases
Solution Approach 1:
The patent dynamically adjusts the resolution level of the 3D model stream based on real-time network conditions and user preferences. When network latency is detected, the system automatically switches to lower resolution levels to reduce transmission time, thereby resolving the contradiction between visual quality and network latency.
Solution Approach 2:
The patent pre-loads and prepares multiple resolution levels of the 3D model in advance, allowing the system to quickly switch between resolution levels based on network conditions without waiting for re-encoding, thereby reducing latency while maintaining quality when possible.
3Device complexity
If uniform resolution degradation is applied, then data transfer is simplified, but visual quality is lost in important regions
Solution Approach 1:
The patent implements non-uniform resolution degradation where different regions of the 3D model are rendered at different resolution levels. Important regions (e.g., foreground objects, areas with detailed textures) maintain high resolution while less important regions use lower resolution, preserving visual quality where it matters most while still reducing overall data transfer.
Solution Approach 2:
The patent segments the 3D model into multiple resolution variants and uses intelligent selection to determine which segments to transmit based on network conditions and viewing context. This allows simplified data management while maintaining high visual quality in important regions through selective transmission of high-resolution segments.
Data Source
AI summary
An adaptive streaming system dynamically streams different parts of a three-dimensional (3D) model at different resolutions to maximize visual detail and quality in response to changing network performance and/or client device rendering performance. The system receives a request to view the 3D model from a particular field-of-view. The system associates different priorities to different parts of the 3D model, selects first Gaussian splats associated with nodes at a first level in a tree based on the first Gaussian splats representing the different parts of the 3D model in the particular field-of-view with a first priority, and selects second Gaussian splats associated with nodes at a second level in the tree structure based on the second Gaussians splats representing the different parts of the 3D model in the particular field-of-view with a second priority. The system streams the selected Gaussian splats in response to the request.