NnD: Diffusion-based Generation of Physically-Nonnegative Objects

TL;DR

This paper introduces NnD, a diffusion-based generative model that efficiently produces physically nonnegative 3D objects like clouds, ensuring physical plausibility and reducing computational costs compared to traditional simulation methods.

Contribution

The paper proposes a novel nonnegative diffusion (NnD) model that enforces non-negativity during generation, enabling realistic and computationally efficient creation of complex physical objects.

Findings

Successfully generated realistic 3D cloud models

Generated clouds align with physical cloud physics trends

Expert perception cannot distinguish generated clouds from real ones

Abstract

Most natural objects have inherent complexity and variability. While some simple objects can be modeled from first principles, many real-world phenomena, such as cloud formation, require computationally expensive simulations that limit scalability. This work focuses on a class of physically meaningful, nonnegative objects that are computationally tractable but costly to simulate. To dramatically reduce computational costs, we propose nonnegative diffusion (NnD). This is a learned generative model using score based diffusion. It adapts annealed Langevin dynamics to enforce, by design, non-negativity throughout iterative scene generation and analysis (inference). NnD trains on high-quality physically simulated objects. Once trained, it can be used for generation and inference. We demonstrate generation of 3D volumetric clouds, comprising inherently nonnegative microphysical fields. Our generated clouds are consistent with cloud physics trends. They are effectively not distinguished as non-physical by expert perception.