Decomposing 2D visual generative models: modularization, key technologies, limitations, and potential works  
      In this blog, I discussed existing visual generation models by decomposing key components for training and developing 2D image generative models including Diffusion Models,
      GANs, and visual auto-regressive models.
      
Neural Information Processing Systems (NeurIPS), 2024 
      [
Paper]
      [
Project Page]
      [
Code]