ReFrame: Layer Caching for Accelerated Inference in Real-Time Rendering

TL;DR

ReFrame introduces a layer caching technique for real-time rendering neural networks, significantly reducing latency by reusing intermediate features, with minimal quality loss across various tasks.

Contribution

This work extends feature caching to real-time rendering, optimizing performance and quality trade-offs in encoder-decoder neural networks.

Findings

Achieves 1.4x average speedup in rendering tasks

Effective caching policies balance quality and performance

Applicable to a range of rendering neural network architectures

Abstract

Graphics rendering applications increasingly leverage neural networks in tasks such as denoising, supersampling, and frame extrapolation to improve image quality while maintaining frame rates. The temporal coherence inherent in these tasks presents an opportunity to reuse intermediate results from previous frames and avoid redundant computations. Recent work has shown that caching intermediate features to be reused in subsequent inferences is an effective method to reduce latency in diffusion models. We extend this idea to real-time rendering and present ReFrame, which explores different caching policies to optimize trade-offs between quality and performance in rendering workloads. ReFrame can be applied to a variety of encoder-decoder style networks commonly found in rendering pipelines. Experimental results show that we achieve 1.4x speedup on average with negligible quality loss in three real-time rendering tasks. Code available: https://ubc-aamodt-group.github.io/reframe-layer-caching/