Frequency-Aware Flow Matching for High-Quality Image Generation

2026-04-16

AI Summary

This work addresses the imbalanced modeling of frequency components in conventional flow matching approaches for image generation, which often leads to insufficient recovery of high-frequency details. To overcome this limitation, the authors propose Frequency-Aware Flow Matching (FreqFlow), the first method to integrate frequency decomposition into the flow matching framework. FreqFlow employs a dual-branch architecture that separately processes low- and high-frequency components and incorporates a time-dependent adaptive weighting mechanism to jointly optimize global structure and local texture in the latent space. Evaluated on class-conditional ImageNet-256 generation, the method achieves an FID of 1.38, substantially outperforming DiT (by −0.79) and SiT (by −0.58), and demonstrates enhanced image coherence and texture sharpness.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Flow matching models have emerged as a powerful framework for realistic image generation by learning to reverse a corruption process that progressively adds Gaussian noise. However, because noise is injected in the latent domain, its impact on different frequency components is non-uniform. As a result, during inference, flow matching models tend to generate low-frequency components (global structure) in the early stages, while high-frequency components (fine details) emerge only later in the reverse process. Building on this insight, we propose Frequency-Aware Flow Matching (FreqFlow), a novel approach that explicitly incorporates frequency-aware conditioning into the flow matching framework via time-dependent adaptive weighting. We introduce a two-branch architecture: (1) a frequency branch that separately processes low- and high-frequency components to capture global structure and refine textures and edges, and (2) a spatial branch that synthesizes images in the latent domain, guided by the frequency branch's output. By explicitly integrating frequency information into the generation process, FreqFlow ensures that both large-scale coherence and fine-grained details are effectively modeled low-frequency conditioning reinforces global structure, while high-frequency conditioning enhances texture fidelity and detail sharpness. On the class-conditional ImageNet-256 generation benchmark, our method achieves state-of-the-art performance with an FID of 1.38, surpassing the prior diffusion model DiT and flow matching model SiT by 0.79 and 0.58 FID, respectively. Code is available at https://github.com/OliverRensu/FreqFlow.

Problem

• flow matching
• frequency components
• image generation
• high-frequency details
• low-frequency structure

Innovation

• frequency-aware
• flow matching
• two-branch architecture
• adaptive weighting
• high-fidelity image generation