Poisson Flow Consistency Training

2025-10-22

AI Summary

Existing Poisson Flow Consistency Models (PFCMs) rely on knowledge distillation from pre-trained PFGM++ models, limiting their generalization to multimodal data. To address this, we propose Poisson Flow Consistency Training (PFCT), the first end-to-end, teacher-free training framework for PFCMs. PFCT introduces three key innovations within the Poisson flow consistency paradigm: (i) a tailored perturbation kernel design, (ii) Beta-distributed noise modeling, and (iii) a sinusoidal discretized time schedule—collectively eliminating reliance on distilled supervision. Evaluated on low-dose CT denoising, PFCT achieves substantial improvements in generation quality (LPIPS reduced by 12.3%, SSIM increased by 0.018), matching state-of-the-art consistency models. This work breaks the dependency of PFCMs on teacher models, significantly broadening their applicability to non-image modalities and resource-constrained deployment scenarios.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The Poisson Flow Consistency Model (PFCM) is a consistency-style model based on the robust Poisson Flow Generative Model++ (PFGM++) which has achieved success in unconditional image generation and CT image denoising. Yet the PFCM can only be trained in distillation which limits the potential of the PFCM in many data modalities. The objective of this research was to create a method to train the PFCM in isolation called Poisson Flow Consistency Training (PFCT). The perturbation kernel was leveraged to remove the pretrained PFGM++, and the sinusoidal discretization schedule and Beta noise distribution were introduced in order to facilitate adaptability and improve sample quality. The model was applied to the task of low dose computed tomography image denoising and improved the low dose image in terms of LPIPS and SSIM. It also displayed similar denoising effectiveness as models like the Consistency Model. PFCT is established as a valid method of training the PFCM from its effectiveness in denoising CT images, showing potential with competitive results to other generative models. Further study is needed in the precise optimization of PFCT and in its applicability to other generative modeling tasks. The framework of PFCT creates more flexibility for the ways in which a PFCM can be created and can be applied to the field of generative modeling.

Problem

• Develops standalone training method for PFCM without pretrained models
• Improves CT image denoising performance using novel noise distribution
• Enables flexible PFCM creation for generative modeling applications

Innovation

• Leveraged perturbation kernel to remove pretrained PFGM++ dependency
• Introduced sinusoidal discretization schedule for adaptability
• Applied Beta noise distribution to improve sample quality