# Data-Driven Synthesis of Smoke Flows with CNN-based Feature Descriptors

## TL;DR

This paper introduces a CNN-based data-driven method for high-resolution smoke flow synthesis that efficiently reuses space-time flow data, capturing small-scale details and flow dynamics with resolution independence.

## Contribution

It presents a novel descriptor learning approach using CNNs and a deformation limiting patch advection method for stable, high-resolution, and resolution-independent smoke flow synthesis.

## Key findings

- High-resolution smoke volumes with detailed small-scale features
- Efficient data reuse through learned descriptors for flow localization
- Stable, non-dissipative flow simulations with natural motion integration

## Abstract

We present a novel data-driven algorithm to synthesize high-resolution flow simulations with reusable repositories of space-time flow data. In our work, we employ a descriptor learning approach to encode the similarity between fluid regions with differences in resolution and numerical viscosity. We use convolutional neural networks to generate the descriptors from fluid data such as smoke density and flow velocity. At the same time, we present a deformation limiting patch advection method which allows us to robustly track deformable fluid regions. With the help of this patch advection, we generate stable space-time data sets from detailed fluids for our repositories. We can then use our learned descriptors to quickly localize a suitable data set when running a new simulation. This makes our approach very efficient, and resolution independent. We will demonstrate with several examples that our method yields volumes with very high effective resolutions, and non-dissipative small scale details that naturally integrate into the motions of the underlying flow.