Diffusion models
Diffusion Models are generative models that work by destroying training data through the successive addition of Gaussian noise, and then learning to recover the data by reversing this noising process. After training, the Diffusion Model can be used to generate data by simply passing randomly sampled noise through the learned denoising process
# Resources
- Diffusion models are autoencoders (Dieleman | Deepmind)
- Guidance: a cheat code for diffusion models (Dieleman | Deepmind)
- High Fidelity Image Generation Using Diffusion Models
- Introduction to deep generative modeling: Diffusion-based Deep Generative Models
- Introduction to Diffusion Models for Machine Learning
- Improving Diffusion Models as an Alternative To GANs
- What are Diffusion Models? - Lilian Weng
- Diffusion models from scratch in PyTorch
- An In-Depth Guide to Denoising Diffusion Probabilistic Models – From Theory to Implementation
# Talks
- #TALK Diffusion Probabilistic Models - MIT, Jascha Sohl-Dickstein
- #TALK What are Diffusion Models? - Ari Seff, math explained
- #TALK Diffusion models explained. How does OpenAI’s GLIDE work? - AI Coffee break with Laetitia
# Code
- #CODE
Diffusers (huggingface)
- Diffusers provides pretrained diffusion models across multiple modalities, such as vision and audio, and serves as a modular toolbox for inference and training of diffusion models
- https://towardsdatascience.com/hugging-face-just-released-the-diffusers-library-846f32845e65
- #TALK https://www.youtube.com/watch?v=UzkdOg7wWmI
# References
- #PAPER Denoising Diffusion Probabilistic Models (Ho 2020)
- #PAPER
Diffusion Models Beat GANs on Image Synthesis (Dhariwal 2021)
- #CODE https://github.com/openai/guided-diffusion
- Diffusion models are a class of likelihood-based models that have shown to produce high-quality images with desired properties such as distribution coverage and easy scalability. These models generate samples by gradually removing noise from a signal.
- The proposed method brings improvements to diffusions models that have worked for GANs, such as improved model architecture and a scheme to trade off diversity for quality. The proposed diffusion model achieves several state-of-the-art results, surpassing GANs on several metrics and datasets
- Paper explained
- #PAPER Improved Denoising Diffusion Probabilistic Models (Nichol 2021)
- #PAPER Cascaded Diffusion Models for High Fidelity Image Generation (Ho 2021)
- #PAPER Classifier-Free Diffusion Guidance (Ho 2022)
- #PAPER #REVIEW Understanding Diffusion Models: A Unified Perspective (Luo 2022)
- #PAPER High-Resolution Image Synthesis with Latent Diffusion Models (Rombach 2022)
- #PAPER Palette: Image-to-Image Diffusion Models (Saharia 2022)