Streaming rendered content is an attractive way to bring high-quality graphics to billions of mobile devices that do not have sufficient rendering power. Existing solutions render content on a server at a fixed frame rate, typically 30 or 60 frames per second, and reduce resolution when bandwidth is restricted. Here, we argue that when streaming graphics content with fast motion, higher quality is achieved when both the frame rate and the resolution are adjusted dynamically based on the content and its motion. We propose a system in which a small neural network predicts the optimal frame rate and resolution for a given transmission bandwidth, content, and motion velocity. This prediction maximizes perceived rendering quality and reduces computational cost under constrained transmission bandwidth. The network is trained on a large dataset of rendered content, which was labeled with a perceptual video quality metric.
