Foveated Animations for Efficient Crowd Simulation

In this paper we focus on exploring the perception of adaptive animation rates on crowd simulations. We propose the concept of foveated animations; adaptive rate of on-screen animations in the viewer’s periphery to save system resources without the user experience being impacted.

We found that the way our peripheral vision reacts to visual crowding where stimuli are grouped and perceived statistically can be heavily exploited when rendering animated crowds. When thousands of characters need bone transformations, blending, and other factors calculated per frame, one can skip these operations without any impact to the user experience if using the appropriate level of foveation.

While work has been done to optimize animation costs, there is yet to be any research on utilizing gaze information to optimize this process without perceived loss of quality.

A more detailed page with videos, source code, and the presentation can be seen here:
Project Page