Is Noise Conditioning Necessary for Denoising Generative Models?

Understand

It is widely believed that noise conditioning is indispensable for denoising diffusion models to work successfully.

• This work challenges this belief.
• Motivated by research on blind image denoising, we investigate a variety of denoising-based generative models in the absence of noise conditioning.
• To our surprise, most models exhibit graceful degradation, and in some cases, they even perform better without noise conditioning.