jeongjiheon/scorediffusionmodel

The Pytorch Tutorial of Score-based and Diffusion Model

Tutorial of Score-based & Diffusion Model from Scratch with PyTorch

• We are going to make a tutorial on models such as NCSN, DDPM, DDIM, VESDE/VPSDE, and LDM.
• I'm going to organize my posts(Korean) on the blog with the code. It will be easier to understand if you look at it with the text.

What is Score-based Diffusion Model?

Content ( Tutorial & Blog (KR) )

Model	Tutorial Code	Blog	Paper	Official Code	Model Type
Generative Modeling by Estimating Gradients of the Data Distribution	Link	Link	Paper	GitHub	Score-based Generative Model
Denoising Diffusion Probabilistic Models	Link	Link	Paper	GitHub	Diffusion Model
Denoising Diffusion Implicit Models	Link		Paper	GitHub	Diffusion Model
Score-based generative model through stochastic differential equation			Paper	GitHub	VESDE / VPSDE
Latent Diffusion Models			Paper	GitHub	Diffusion + AE

Content ( only Blog (KR) )

Model	Blog	Paper	Official Code	Model Type
Improved Techniques for Training Score-Based Generative Models	Link	Paper	GitHub	Score-based Generative Model
Improved Denoising Diffusion Probabilistic Models		Paper	GitHub	Diffusion Model
Diffusion Models Beat GANs on Image Synthesis		Paper	GitHub	Diffusion Model
CLASSIFIER-FREE DIFFUSION GUIDANCE		Paper	-	Diffusion Model

Contributor

Name	Github	E-mail	Linkedin
Jiheon, Jeong	[GitHub]	[E-mail]	[Linkedin]
Yujin, Nam	[GitHub]	[E-mail]	[Linkedin]

scorediffusionmodel's Issues

What does components in NCSN_MNIST means?

When I looked through your code of NCSN_MNIST.py, I'm wondering that ConditionalBatchNorm2d/ConditionalInstanceNorm2d/CRPBlock/CondCRPBlock/CondRCUBlock/CondMSFBlock/CondRefineBlock/ConvMeanPool/MeanPoolConv/ConditionalResidualBlock/CondRefineNetDilated and so on.

Actually, I'm not able to understand what prefix such as "Cond", "Cons", "CRP", "RCU", "MSF".
In this regard, I would like you to let me know these prefixes.