Summary of Pagoda: Progressive Growing Of a One-step Generator From a Low-resolution Diffusion Teacher, by Dongjun Kim et al.
PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher
by Dongjun Kim, Chieh-Hsin Lai, Wei-Hsiang Liao, Yuhta Takida, Naoki Murata, Toshimitsu Uesaka, Yuki Mitsufuji, Stefano Ermon
First submitted to arxiv on: 23 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Progressive Growing of Diffusion Autoencoder (PaGoDA) pipeline reduces the computational cost of training diffusion models while maintaining their performance. By training a diffusion model on downsampled data, distilling the pretrained model, and progressively super-resolving it, PaGoDA achieves state-of-the-art results on ImageNet across various resolutions and tasks. This approach can be applied directly to latent space, enabling compression alongside pre-trained autoencoders in Latent Diffusion Models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary PaGoDA is a new way to make computer-generated images faster and better. It works by first training an image model on small, low-quality pictures, then using that model to help train an even better version of itself. This process is repeated multiple times until the final, high-quality model is achieved. The result is a much more efficient and effective way to generate realistic images. |
Keywords
» Artificial intelligence » Autoencoder » Diffusion » Diffusion model » Latent space
