Greedy Growing Enables High-Resolution Pixel-Based Diffusion Models

• URL: http://arxiv.org/abs/2405.16759v1
• Date: Mon, 27 May 2024 02:12:39 GMT
• Title: Greedy Growing Enables High-Resolution Pixel-Based Diffusion Models
• Authors: Cristina N. Vasconcelos, Abdullah Rashwan Austin Waters, Trevor Walker, Keyang Xu, Jimmy Yan, Rui Qian, Shixin Luo, Zarana Parekh, Andrew Bunner, Hongliang Fei, Roopal Garg, Mandy Guo, Ivana Kajic, Yeqing Li, Henna Nandwani, Jordi Pont-Tuset, Yasumasa Onoe, Sarah Rosston, Su Wang, Wenlei Zhou, Kevin Swersky, David J. Fleet, Jason M. Baldridge, Oliver Wang,
• Abstract summary: We propose a greedy algorithm that grows the architecture into high-resolution end-to-end models. This enables a single stage model capable of generating high-resolution images without the need of a super-resolution cascade. Our results rely on public datasets and show that we are able to train non-cascaded models up to 8B parameters with no further regularization schemes.
• Score: 41.67994377132345
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We address the long-standing problem of how to learn effective pixel-based image diffusion models at scale, introducing a remarkably simple greedy growing method for stable training of large-scale, high-resolution models. without the needs for cascaded super-resolution components. The key insight stems from careful pre-training of core components, namely, those responsible for text-to-image alignment {\it vs.} high-resolution rendering. We first demonstrate the benefits of scaling a {\it Shallow UNet}, with no down(up)-sampling enc(dec)oder. Scaling its deep core layers is shown to improve alignment, object structure, and composition. Building on this core model, we propose a greedy algorithm that grows the architecture into high-resolution end-to-end models, while preserving the integrity of the pre-trained representation, stabilizing training, and reducing the need for large high-resolution datasets. This enables a single stage model capable of generating high-resolution images without the need of a super-resolution cascade. Our key results rely on public datasets and show that we are able to train non-cascaded models up to 8B parameters with no further regularization schemes. Vermeer, our full pipeline model trained with internal datasets to produce 1024x1024 images, without cascades, is preferred by 44.0% vs. 21.4% human evaluators over SDXL.

Related papers

• Hierarchical Patch Diffusion Models for High-Resolution Video Generation [50.42746357450949]
  We develop deep context fusion, which propagates context information from low-scale to high-scale patches in a hierarchical manner. We also propose adaptive computation, which allocates more network capacity and computation towards coarse image details. The resulting model sets a new state-of-the-art FVD score of 66.32 and Inception Score of 87.68 in class-conditional video generation.
  arXiv  Detail & Related papers  (2024-06-12T01:12:53Z)
• FouriScale: A Frequency Perspective on Training-Free High-Resolution Image Synthesis [48.9652334528436]
  We introduce an innovative, training-free approach FouriScale from the perspective of frequency domain analysis. We replace the original convolutional layers in pre-trained diffusion models by incorporating a dilation technique along with a low-pass operation. Our method successfully balances the structural integrity and fidelity of generated images, achieving an astonishing capacity of arbitrary-size, high-resolution, and high-quality generation.
  arXiv  Detail & Related papers  (2024-03-19T17:59:33Z)
• Matryoshka Diffusion Models [38.26966802461602]
  Diffusion models are the de facto approach for generating high-quality images and videos. We introduce Matryoshka Diffusion Models, an end-to-end framework for high-resolution image and video synthesis. We demonstrate the effectiveness of our approach on various benchmarks, including class-conditioned image generation, high-resolution text-to-image, and text-to-video applications.
  arXiv  Detail & Related papers  (2023-10-23T17:20:01Z)
• PixelPyramids: Exact Inference Models from Lossless Image Pyramids [58.949070311990916]
  Pixel-Pyramids is a block-autoregressive approach with scale-specific representations to encode the joint distribution of image pixels. It yields state-of-the-art results for density estimation on various image datasets, especially for high-resolution data. For CelebA-HQ 1024 x 1024, we observe that the density estimates are improved to 44% of the baseline despite sampling speeds superior even to easily parallelizable flow-based models.
  arXiv  Detail & Related papers  (2021-10-17T10:47:29Z)
• Spatial-Separated Curve Rendering Network for Efficient and High-Resolution Image Harmonization [59.19214040221055]
  We propose a novel spatial-separated curve rendering network (S$2$CRNet) for efficient and high-resolution image harmonization. The proposed method reduces more than 90% parameters compared with previous methods. Our method can work smoothly on higher resolution images in real-time which is more than 10$times$ faster than the existing methods.
  arXiv  Detail & Related papers  (2021-09-13T07:20:16Z)
• InfinityGAN: Towards Infinite-Resolution Image Synthesis [92.40782797030977]
  We present InfinityGAN, a method to generate arbitrary-resolution images. We show how it trains and infers patch-by-patch seamlessly with low computational resources.
  arXiv  Detail & Related papers  (2021-04-08T17:59:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.