Interactive Isosurface Visualization in Memory Constrained Environments Using Deep Learning and Speculative Raycasting.

New web technologies have enabled the deployment of powerful GPU-based computational pipelines that run entirely in the web browser, opening a new frontier for accessible scientific visualization applications. However, these new capabilities do not address the memory constraints of lightweight end-user devices encountered when attempting to visualize the massive data sets produced by today's simulations and data acquisition systems. We propose a novel implicit isosurface rendering algorithm for interactive visualization of massive volumes within a small memory footprint.