Frequency-Aware Reconstruction of Fluid Simulations with Generative Networks

TL;DR

This paper introduces a frequency-aware loss function for generative networks to improve the reconstruction quality of fluid simulations, especially in high-frequency details, resulting in more perceptually accurate results without extra training time.

Contribution

The paper proposes a novel frequency-aware loss function that enhances the reconstruction of high-frequency details in fluid simulations using generative networks.

Findings

Improved reconstruction quality in mid-frequency bands.

Perceptually better results with comparable training time.

Enhanced high-frequency detail preservation.

Abstract

Convolutional neural networks were recently employed to fully reconstruct fluid simulation data from a set of reduced parameters. However, since (de-)convolutions traditionally trained with supervised L1-loss functions do not discriminate between low and high frequencies in the data, the error is not minimized efficiently for higher bands. This directly correlates with the quality of the perceived results, since missing high frequency details are easily noticeable. In this paper, we analyze the reconstruction quality of generative networks and present a frequency-aware loss function that is able to focus on specific bands of the dataset during training time. We show that our approach improves reconstruction quality of fluid simulation data in mid-frequency bands, yielding perceptually better results while requiring comparable training time.