Data Reduction and Visualization Synthesis via Neural Field Representation

Abstract: Neural field representation has emerged as a trending topic in scientific computing and data visualization. In this talk, I will present two of our recent works on data reduction and visualization synthesis. The first work leverages the multiscale Laplacian pyramid structure and employs multiple small multilayer perceptrons to fit local content or residual blocks at each scale. This leads to a highly compressive neural representation of time-varying volumetric datasets. The second work utilizes a hybrid radiance field representation to efficiently train and infer rendering images given a sparse set of labeled image samples. This enables high-quality visualization synthesis across novel viewpoints, timesteps, isovalues, transfer functions, or simulation parameters. Finally, I will discuss future perspectives for this vibrant research direction.

Bio: Chaoli Wang is a Professor in the Department of Computer Science and Engineering at the University of Notre Dame. He holds a Ph.D. degree in Computer and Information Science from The Ohio State University.