ShaderTransformer: Predicting Shader Quality via One-shot Embedding for Fast Simplification
Event Type
Technical Paper
Interest Areas
Research & Education
Presentation Types
Registration Categories
Full Conference Supporter
Full Conference
Virtual Conference Supporter
Virtual Conference
Exhibitor Additional Full Conference
Exhibitor Full Conference
DescriptionThis paper introduces a multimodal self-attention network, ShaderTransformer, to predict shader variants’ quality for fast shader simplification. Once trained, the network allows one-shot embedding of all variants of the target shader for a specific scene configuration. The positions of the embedded variants represent the simplification errors, which enable fast searching.