Please enable JavaScript.
Coggle requires JavaScript to display documents.
SDXL: Improving Latent Diffusion Models for High-Resolution Image…
SDXL: Improving Latent Diffusion Models for
High-Resolution Image Synthesis
SDXL, a latent diffusion model for text-to-image synthesis
Latent Diffusion Model (LDM)
High-Resolution Image Synthesis with Latent Diffusion Models
diffusion
models (DMs)
equally weighted sequence of denoising autoencoders
latent space
unconditional image synthesis
class-condition
Perceptual Compression
KL-reg
VQ-reg
time-conditional UNet
Conditioning Mechanisms
conditional denoising autoencoder
domain specidic encoder
VAE
Dataset
LSUN
Churches
Bedrooms
FFHQ
CelebA-HQ
ImageNet
Conceptual Captions
LAION
Tasks
Super-Resolution
inpainting
layout-condition
text-to-image
unconditional-image
Comparison
FID
Precision-and-Recall
GANs
LSGM
DDPM
解析
Latent Diffusion Models论文解读
cross-attention
refinement model
Tasks
3D classification
controllable image editing
image personalization
synthetic data augmentation
graphical user interface prototyping
music generation
reconstructing images from fMRI brain scans
image-to-image
Future work
Single stage
Text synthesis
Architecture
Distillation
Not require noise-schedule corrections
OpenCLIP ViT-bigG
CLIP ViT-L
apparent resolution
sizeconditioning
Fourier feature encoding
Model
CIN-512-only
CIN-nocond
CIN-size-cond
SDXL-VAE
SD-VAE 1.x
SD-VAE 2.x
DDIM
Performance
FID
IS
CLIP
random cropping
cfg-scale 8.0
conditioning-augmentation
COCO2017
Stochastic Differential Editing (SDEdit)
SDEDIT: GUIDED IMAGE SYNTHESIS AND EDITING
WITH STOCHASTIC DIFFERENTIAL EQUATIONS
stochastic differential equation (SDE)
Score-based generative modeling through stochastic differential equations
faithfulness
realism
GANs inversion and editing
Conditional GANs
stroke-based image synthesis
stroke-based image editing
image compositing
ordinary differential equations (ODEs)
Gaussian distribution
Variance Exploding SDE (VE-SDE)
Variance Preserving (VP)
Dataset
LSUN
CelebA
FFHQ
Performance
Kernel Inception Score (KID)
Mechanical Turk (MTurk)
LPIPS
Comparison
StyleGAN2ADA
Training generative adversarial networks with limited data
In-domain GAN inversion
In-domain gan inversion for real image editing
e4e
SC-FEGAN
Probability Flow
ordinary differential equation (ODE)
score function
stochastic differential equation (SDE)
standard Wiener process
Langevin diffusion component
denoiser
denoising score matching (DSM)
Classifier-free guidance
guidance strength
Comparison
DeepFloyd IF
DALLE-2
Bing Image Creator
Midjourney v5.2
Parti (P2) prompts
Seed 3