Multi-Scale Generative Modeling with Heat Dissipation Flow Matching

TL;DR

This paper introduces Heat Dissipation Flow Matching (HDFM), a novel ODE-based generative model integrating multi-scale priors via a heat-dissipation process, improving image generation quality.

Contribution

HDFM uniquely combines heat dissipation processes with flow matching, addressing ill-posedness and high-dimensional regression in blur-based diffusion models.

Findings

HDFM outperforms baseline methods on multiple datasets.

Incorporating heat dissipation improves multi-scale detail preservation.

x-prediction mitigates high-dimensional regression challenges.

Abstract

Diffusion models are widely used in image generation, with most relying on noise-based corruption and denoising. A distinct branch instead uses blur as the main corruption, preserving better color budgets and multi-scale detail by providing multi-scale priors. However, blur-based models remain in SDE-based frameworks and are not integrated into ODE-based frameworks, such as Flow Matching (FM). Meanwhile, in the blur-based formulation, the classical inverse heat-dissipation (IHD) process faces an ill-posed challenge. Moreover, under the data-manifold assumption, regressing blurred images from high-dimensional noise (or velocity) space is also difficult. We propose Heat Dissipation Flow Matching (HDFM), which introduces a continuous blurred (heat-dissipation) process into FM to inject multi-scale priors. HDFM aligns an interpolated heat-dissipation path to address ill-posedness and adopts $x$-prediction to mitigate high-dimensional regression difficulty. Toy experiments and ablation studies show that HDFM consistently benefits from both blur and $x$-prediction. The performance of HDFM outperforms most baseline methods on all datasets.