Texture Mip Prediction for Pop-Free Game Streaming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing texture streaming methods in video games suffer from latency issues and texture 'popping' due to progressive loading of higher quality mips, which detracts from the realism and immersion of the game graphics.
Innovation Solution
Implementing a machine learning model to predict higher quality mips from lower quality mips, allowing immediate output of predicted textures until the actual higher quality mips are loaded, thereby reducing latency and eliminating texture popping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If higher quality mip levels are loaded progressively from storage media, then memory usage is optimized and latency is reduced, but texture quality transitions cause visible popping artifacts
Solution Approach 1:
The system performs preliminary action by loading and caching lower quality mip levels into memory before they are actually needed for rendering. This allows the texture streamer to immediately provide higher quality mip levels from memory when requested, eliminating the progressive loading delay and associated popping artifacts while maintaining optimized memory usage through selective caching.
2Reliability
If higher priority textures are transferred before lower priority textures, then critical visual quality is maintained, but lower priority textures experience increased loading delay
Solution Approach 1:
The system applies preliminary action by pre-loading and caching lower priority textures at lower mip levels into memory during idle periods or alongside higher priority texture loading. When higher quality versions of these lower priority textures are subsequently requested, they can be quickly synthesized or loaded from the cached base versions, reducing the delay penalty of priority-based loading while maintaining visual quality where needed.
Data Source
AI summary
A data processing apparatus comprising circuitry configured to: obtain a first version of a texture having a first quality level; generate, using the first version of the texture, a prediction of a second version of the texture having a second quality level, the second quality level being higher in quality than the first quality level; and output, for display, the prediction of the second version of the texture.