\textbf{Naresh Kumar Devulapally}
\text{CSE 4/555: Intro to Pattern Recognition}
\text{Lecture 4}
\text{Latent Diffusion Models and latest architectures}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Diffusion Models - Part 3}
  • Recap of the VAE Architecture
  • Recap of the Pixel Level Diffusion Model
  • Conditional Diffusion Model
  • Classifier Guidance v/s Classifier Free Guidance
  • Why Latent Diffusion Models?
  • Latent Diffusion Models (LDMs) explained
  • Cross Attention in LDMs.
  • Diffusion Models for various Computer Vision tasks.
  • Tips to complete capstone project milestone 2.
  • Information about Guest Talk on July 31 2025.

\( \text{Agenda of this Lecture:}\)

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{VAEs v/s Diffusion Models}

Gaussian Variable

Gaussian Variable

\( \mathcal{L}_{\text{VAE}}  = \text{Reconstruction} + \text{Prior Matching} \)

\( \mathcal{L}_{\text{Diff}}  = \text{Reconstruction} + \text{Prior Matching} + \text{Noise Matching} \)

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{VAEs v/s Diffusion Models}

Gaussian Variable

Gaussian Variable

\( \mathcal{L}_{\text{VAE}}  = \text{Reconstruction} + \text{Prior Matching} \)

\( \mathcal{L}_{\text{Diffusion-Training}}  = \text{Noise Matching} \)

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Diffusion Models}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Diffusion Models}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Diffusion Models}

Unconditional Image Generation

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Conditional Img. Gen. in Diffusion Models}

Classifier Guidance

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Conditional Img. Gen. in Diffusion Models}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{What is the Diffusion Model?}

UNet

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Latent Diffusion Models}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Latent Diffusion Models}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Latent Diffusion Models}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Latent Diffusion Models}
\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Latent Diffusion Models}

Cross Attention Maps

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}
\text{July 10, 2025}
\text{Latent Diffusion Models}

Cross Attention Maps for Editing

\text{Naresh Kumar Devulapally}
\text{Apr. 9, 2026}
\text{CSE 4/555: Pattern Recognition, Sp. 26}

Lecture 4: Latent Diffusion Models and latest architectures

By Naresh Kumar Devulapally

Lecture 4: Latent Diffusion Models and latest architectures

  • 8