Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusion

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hoogeboom, Emiel, Mensink, Thomas, Heek, Jonathan, Lamerigts, Kay, Gao, Ruiqi, Salimans, Tim
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912288331530240
author Hoogeboom, Emiel
Mensink, Thomas
Heek, Jonathan
Lamerigts, Kay
Gao, Ruiqi
Salimans, Tim
author_facet Hoogeboom, Emiel
Mensink, Thomas
Heek, Jonathan
Lamerigts, Kay
Gao, Ruiqi
Salimans, Tim
contents Latent diffusion models have become the popular choice for scaling up diffusion models for high resolution image synthesis. Compared to pixel-space models that are trained end-to-end, latent models are perceived to be more efficient and to produce higher image quality at high resolution. Here we challenge these notions, and show that pixel-space models can be very competitive to latent models both in quality and efficiency, achieving 1.5 FID on ImageNet512 and new SOTA results on ImageNet128, ImageNet256 and Kinetics600. We present a simple recipe for scaling end-to-end pixel-space diffusion models to high resolutions. 1: Use the sigmoid loss-weighting (Kingma & Gao, 2023) with our prescribed hyper-parameters. 2: Use our simplified memory-efficient architecture with fewer skip-connections. 3: Scale the model to favor processing the image at a high resolution with fewer parameters, rather than using more parameters at a lower resolution. Combining these with guidance intervals, we obtain a family of pixel-space diffusion models we call Simpler Diffusion (SiD2).
format Preprint
id arxiv_https___arxiv_org_abs_2410_19324
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusion
Hoogeboom, Emiel
Mensink, Thomas
Heek, Jonathan
Lamerigts, Kay
Gao, Ruiqi
Salimans, Tim
Computer Vision and Pattern Recognition
Machine Learning
Latent diffusion models have become the popular choice for scaling up diffusion models for high resolution image synthesis. Compared to pixel-space models that are trained end-to-end, latent models are perceived to be more efficient and to produce higher image quality at high resolution. Here we challenge these notions, and show that pixel-space models can be very competitive to latent models both in quality and efficiency, achieving 1.5 FID on ImageNet512 and new SOTA results on ImageNet128, ImageNet256 and Kinetics600. We present a simple recipe for scaling end-to-end pixel-space diffusion models to high resolutions. 1: Use the sigmoid loss-weighting (Kingma & Gao, 2023) with our prescribed hyper-parameters. 2: Use our simplified memory-efficient architecture with fewer skip-connections. 3: Scale the model to favor processing the image at a high resolution with fewer parameters, rather than using more parameters at a lower resolution. Combining these with guidance intervals, we obtain a family of pixel-space diffusion models we call Simpler Diffusion (SiD2).
title Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusion
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2410.19324