S3RP: Self-Supervised Super-Resolution and Prediction for Advection-Diffusion Process

TL;DR

This paper introduces S3RP, a self-supervised super-resolution model for advection-diffusion processes that does not require high-resolution ground-truth data, using physics-based regularizations and a recurrent Wasserstein autoencoder to handle uncertainty.

Contribution

The proposed model enables super-resolution without HR ground-truth data by combining physics-based regularizations with a recurrent autoencoder to model uncertainty.

Findings

Successfully reconstructs high-resolution data without HR ground-truth.

Employs physics-based regularizations to guide super-resolution.

Models uncertainty effectively with a recurrent Wasserstein autoencoder.

Abstract

We present a super-resolution model for an advection-diffusion process with limited information. While most of the super-resolution models assume high-resolution (HR) ground-truth data in the training, in many cases such HR dataset is not readily accessible. Here, we show that a Recurrent Convolutional Network trained with physics-based regularizations is able to reconstruct the HR information without having the HR ground-truth data. Moreover, considering the ill-posed nature of a super-resolution problem, we employ the Recurrent Wasserstein Autoencoder to model the uncertainty.